GPT-6 Astra: score and benchmark evidence
Historical release · · OpenAI
140.8 points · Rank 1 among 19 ranked models · Stability range 130.1–151.5 points.
Rank follows the point estimate; it does not establish statistically significant superiority. The scale is anchored at mean 100, SD 15 in the frozen calibration cohort, not human IQ or a percentage.
One score summarizes demonstrated capability across published reasoning settings and seven equally weighted domains. Related benchmark variants share a family budget, and influence from one evaluator is capped.
Release tli-2026-v1.0-2026-09-10-benchmarks-1. Methodology and limitations for this release.
This page preserves historical evidence. See the current model record.
Capability domains
Domain scores use calibration-cohort standard-deviation units. Unmeasured domains are shown as unavailable, not zero.
| Domain | Score (SD units) | Observed benchmarks |
|---|---|---|
| Knowledge & reasoning | 1.47 | 16 |
| Coding & software engineering | 1.23 | 11 |
| Agentic tool & computer use | 1.30 | 7 |
| Professional & real-world work | 0.61 | 13 |
| Multimodal & vision | 1.10 | 4 |
| Cybersecurity | 1.01 | 1 |
| Preference & communication | 1.06 | 1 |
Published benchmark evidence
53 results contribute to this score, from 83 collected results and 44 contributing benchmark families. Counts are not independent sample sizes.
Results, source citations, review decisions and compatible reasoning modes are stored in a versioned relational database. Each publication uses a sealed, checksum-verified export; CSV files are optional exports, not editable scoring inputs. Previous revisions remain available for audit. Collected means a numeric result exists in the database. Used means its source, exact model checkpoint and benchmark protocol have been reviewed and its benchmark has a frozen calibration supported by multiple providers. For each benchmark we select the highest published aggregate across comparable reasoning modes, not necessarily maximum effort, and count it once. More tested settings provide more chances to record a high score; this selection advantage is not corrected by the conditional interval. Developer-reported and sole/unspecified settings remain eligible and disclosed. Unresolved does not mean false or useless: conflicting metrics, unverified citations and unmatched versions stay collected until reconciled.
The score stability range is a 95% bootstrap score interval conditional on frozen calibration and comparable evidence availability. It resamples benchmark families and evaluators and adds a frozen missing-evidence correction. It excludes selection across reasoning settings, selective publication, calibration changes and new domains. It is not a 95% guarantee about true capability or where a future score will land. Sparse or conflicting evidence can widen the range.
Reporting sensitivity: ±1.2 points. Sensitivity to selectively published results. This stress range is not a confidence interval or a probability statement.
GDPval-AA v2 (Elo): 1580.2 native-points
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max); Artificial Analysis published evaluation harness
Metric: Elo
Original source · Retrieved 2026-09-10
Exact checkpoint, metric and published aggregate matched to the evaluator source. Date is retrieval date; first publication date is not inferred.
Terminal-Bench 2.1: 89.88764%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (high); Artificial Analysis published evaluation harness
Metric: pass@1, AA harness
Original source · Retrieved 2026-09-06
Best published result among 6 reasoning configurations of the exact AA model release; fallback configurations excluded.
DeepSWE v1.1: 74.115044%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: gpt-6-astra; xhigh; mini-swe-agent
Metric: pass@1 over scored rollout attempts
Original source · Retrieved 2026-09-06
Selected the best pass@1 across published reasoning efforts within mini-swe-agent. Pass@4 is not substituted.
AA Intelligence Index v4.1.1 (score): 61.2 native-points
Not used in the score. Source category not recorded. Review status: excluded.
Held-out composite, never a scoring input
GPQA Diamond: 96.262626%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (xhigh); Artificial Analysis published evaluation harness
Metric: accuracy, AA
Original source · Retrieved 2026-09-06
Best published result among 6 reasoning configurations of the exact AA model release; fallback configurations excluded.
AA-LCR: 74.3%
Not used in the score. Source category: independent. Review status: unresolved.
Configuration: GPT-6 Astra (max)
Original source · Retrieved 2026-09-06
Current AA result is LCR v1.1. The inherited AA-LCR row has no reviewed version match; a new version cannot silently substantiate the older value.
Finance Agent v2 — Vals: 53.54%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: openai_gpt-6-astra; Vals max effort
Metric: fabv2 overall percentage in Vals benchmark harness
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
Excel Modeling Benchmark — Vals: 71.7%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: openai_gpt-6-astra; Vals max effort
Metric: emb overall percentage in Vals benchmark harness
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
SciCode: 56.481481%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max); Artificial Analysis published evaluation harness
Metric: accuracy, AA harness
Original source · Retrieved 2026-09-06
Best published result among 6 reasoning configurations of the exact AA model release; fallback configurations excluded.
Humanity's Last Exam (no tools): 54.680259%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max); Artificial Analysis published evaluation harness
Metric: accuracy, no tools, AA
Original source · Retrieved 2026-09-06
Best published result among 6 reasoning configurations of the exact AA model release; fallback configurations excluded.
Legal Research Bench — Vals: 39.42%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: openai_gpt-6-astra; Vals max effort
Metric: legal_research overall percentage in Vals benchmark harness
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
Harvey LAB (Vals): 5.42%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: openai_gpt-6-astra; Vals max effort
Metric: hlab overall percentage in Vals benchmark harness
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
AA-Briefcase Overall Elo: 1562.01 native-points
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max); Artificial Analysis published evaluation harness
Metric: overall Elo
Original source · Retrieved 2026-09-10
Exact checkpoint, metric and published aggregate matched to the evaluator source. Date is retrieval date; first publication date is not inferred.
MedScribe — Vals: 87.91%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: openai_gpt-6-astra; Vals max effort
Metric: medscribe overall percentage in Vals benchmark harness
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
MedCode — Vals: 48.49%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: openai_gpt-6-astra; Vals max effort
Metric: medcode overall percentage in Vals benchmark harness
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
AA-Omniscience: 43.733333 native-points
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (high); Artificial Analysis published evaluation harness
Metric: knowledge reliability index -100 to 100
Original source · Retrieved 2026-09-06
Best published result among 6 reasoning configurations of the exact AA model release; fallback configurations excluded.
SAGE — Vals: 46.37%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: openai_gpt-6-astra; Vals max effort
Metric: sage overall percentage in Vals benchmark harness
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
Code Migration †: 67.74%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: openai_gpt-6-astra; Vals max effort
Metric: code-migration overall percentage in Vals benchmark harness
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
Humanity's Last Exam w/ tools: 57.2%
Not used in the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; OpenAI comparison table; maximum across reasoning modes
Metric: Humanity's Last Exam (w/ tools)
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
ProgramBench v1 (Raw Pass Rate) — Vals: 85.417%
Contributes to the score. Source category: independent. Review status: reviewed-archive.
Configuration: openai/gpt-6-astra · reasoning_effort=max
Metric: Raw Pass Rate: mean percent of hidden behavioral tests passed per task. Not Fully Resolved (all tests pass) and not Almost Resolved (≥95% of tests pass).
Original source · Retrieved 2026-09-05
Matched exact metric/value and the archived benchmark-specific configuration review. Unknown native effort stays explicitly unknown.
FrontierMath v2 Tier 4: 97.560976%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: gpt-6-astra_medium; task version 2.0.0
Metric: mean score percentage, FrontierMath-Tier-4-v2-Private
Original source · Retrieved 2026-09-06
Exact Epoch task version; selected best mean score across reported reasoning configurations. No pooling of task versions or substitution of best scorer output.
APEX-Agents Mean Criteria Passed: 75.1%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: gpt-6-astra-max; max; loop_truncated_tools_agent
Metric: mean-score
Original source · Retrieved 2026-09-10
Same 240-task, 31-world APEX benchmark and mean-score metric. Current published repeated-run aggregate (959 samples); pass@1 is not substituted.
BrowseComp: 91.5%
Contributes to the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; OpenAI comparison table; maximum across reasoning modes
Metric: BrowseComp
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
MMMU-Pro: 86.878613%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max); Artificial Analysis published evaluation harness
Metric: accuracy, AA
Original source · Retrieved 2026-09-06
Best published result among 6 reasoning configurations of the exact AA model release; fallback configurations excluded.
ARC-AGI-2: 95%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra; Max; ARC Prize standard harness
Metric: semi-private verified score percentage
Original source · Retrieved 2026-09-06
Best reasoning variant within the verified standard-harness summary. Provider Adapter, public tasks and community agent scores are not substituted.
ExploitBench: 100%
Not used in the score. Source category not recorded. Review status: excluded.
Inherited comparator metric/scaffold equivalence to Astra Cap Percent is not established; the source flags possible contamination. Excluded pending protocol audit.
FrontierCode v1.1 Extended: 64.48%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra; max; codex
Metric: v1.1 Extended cost-adjusted new_score, 150 tasks
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
FrontierCode v1.1 Main — Anthropic H2H: 53.3%
Not used in the score. Source category not recorded. Review status: unresolved.
Not admitted after the September 6 model review: the collected 53.3 for GPT-6 Astra on FrontierCode v1.1 Main — Anthropic H2H still lacks a matched source entry establishing its checkpoint and benchmark-specific metric, tools and retry setup. This is a citation/comparability gap, not a finding that the result is false or useless. More or maximum reasoning effort is not required.
AutomationBench — Anthropic H2H: 41.4%
Not used in the score. Source category not recorded. Review status: unresolved.
Not admitted after the September 6 model review: the collected 41.4 for GPT-6 Astra on AutomationBench — Anthropic H2H still lacks a matched source entry establishing its checkpoint and benchmark-specific metric, tools and retry setup. This is a citation/comparability gap, not a finding that the result is false or useless. More or maximum reasoning effort is not required.
τ³-Banking: 43.092784%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (xhigh); Artificial Analysis published evaluation harness
Metric: strict task success, AA
Original source · Retrieved 2026-09-06
Best published result among 6 reasoning configurations of the exact AA model release; fallback configurations excluded.
CritPt: 31.714286%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max); Artificial Analysis published evaluation harness
Metric: accuracy, AA official grading
Original source · Retrieved 2026-09-06
Best published result among 6 reasoning configurations of the exact AA model release; fallback configurations excluded.
Terminal-Bench 4.0: 58.2%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra; max; Codex
Metric: resolution rate, v4.0.0 66 tasks
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
ProofBench v1.1 — Vals: 99%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: openai_gpt-6-astra; Vals max effort
Metric: proof_bench overall percentage in Vals benchmark harness
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
Vibe Code Bench v1.1 — Vals: 89.59%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: openai_gpt-6-astra; Vals max effort
Metric: vibe-code overall percentage in Vals benchmark harness
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
ARC-AGI-1: 98.5%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra; XHigh; ARC Prize standard harness
Metric: semi-private verified score percentage
Original source · Retrieved 2026-09-06
Best reasoning variant within the verified standard-harness summary. Provider Adapter, public tasks and community agent scores are not substituted.
ARC-AGI-3 (standard harness): 62.71%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra; Max; ARC Prize standard harness
Metric: semi-private verified score percentage
Original source · Retrieved 2026-09-06
Best reasoning variant within the verified standard-harness summary. Provider Adapter, public tasks and community agent scores are not substituted.
BenchCAD (Vision2Code subset): 95.9%
Not used in the score. Developer-reported. Review status: unresolved.
Configuration: OpenAI full Vision2Code set; not Anthropic subset
Original source · Retrieved 2026-09-06
System card identifies the OpenAI comparison as the full 17,900-file set, while this row is a modified 1000-file subset. Requires a separate protocol row.
Terminal-Bench Science 0.1: 64.6%
Contributes to the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: Terminal-Bench Science 0.1 as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
SRE-Bench (pass@1): 88%
Contributes to the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: SRE-Bench (pass@1) as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
OSWorld 2.0 — OpenAI H2H (offline subset): 72.6%
Contributes to the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: OSWorld 2.0 — OpenAI H2H (offline subset) as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
Agents' Last Exam — OpenAI H2H: 59.3%
Contributes to the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: Agents' Last Exam — OpenAI H2H as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
ScreenSpot-Pro: 92.7%
Not used in the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: ScreenSpot-Pro as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
HealthBench: 59.7%
Not used in the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; OpenAI published HealthBench research evaluation; effort unspecified
Metric: raw physician rubric score, not length-adjusted
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
HealthBench Hard: 37.8%
Not used in the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; OpenAI published HealthBench research evaluation; effort unspecified
Metric: raw physician rubric score, not length-adjusted
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
HealthBench Professional (length-adjusted): 63.4%
Contributes to the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: HealthBench Professional (length-adjusted) as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
GeneBench Pro: 37.1%
Not used in the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: GeneBench Pro as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
LifeSciBench: 60.3%
Not used in the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: LifeSciBench as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
SEC-Bench Pro: 85.4%
Not used in the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: SEC-Bench Pro as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
ExploitGym: 42.4%
Not used in the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: ExploitGym as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
AA Coding Agent Index v1.4 (score): 67 native-points
Not used in the score. Source category not recorded. Review status: excluded.
Held-out composite, never a scoring input
Internal Database Migration Tasks — OpenAI: 63.9%
Contributes to the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: Internal Database Migration Tasks — OpenAI as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
Internal Design Tasks — OpenAI: 50%
Contributes to the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: Internal Design Tasks — OpenAI as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
Internal Data Science Tasks — OpenAI: 40.9%
Contributes to the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: Internal Data Science Tasks — OpenAI as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
OpenScore String Quartets (1 - OMR-NED): 0.84 native-points
Not used in the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: OpenScore String Quartets (1 - OMR-NED) as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
MedChemBench (internal): 49.3%
Not used in the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: MedChemBench (internal) as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
MRCR v2 (8 needles, 512K-1M): 96.3%
Not used in the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: MRCR v2 (8 needles, 512K-1M) as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
ExploitBench (Jun-Aug 2026, 300-turn limit): 39%
Not used in the score. Developer-reported. Review status: reviewed-developer.
Configuration: GPT-6 Astra; publisher-reported maximum across reasoning efforts; OpenAI research/API evaluation
Metric: ExploitBench (Jun-Aug 2026, 300-turn limit) as labelled in OpenAI comparison table
Original source · Retrieved 2026-09-06
Developer-reported maximum across reasoning modes. Exact selected mode is not individually disclosed. Research prompts/tools may differ from production; protocol-specific rows remain separate.
Mystery Game Puzzles v1.0.4 — Epoch: 84%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: gpt-6-astra_max; task version 1.0.4
Metric: mean score percentage, Mystery Game Puzzles
Original source · Retrieved 2026-09-06
Exact Epoch task version; selected best mean score across reported reasoning configurations. No pooling of task versions or substitution of best scorer output.
WANDR — Perplexity: 0.682 native-points
Not used in the score. Source category not recorded. Review status: excluded.
Exact reported metric and evaluation configuration unresolved.
FrontierMath Erdos (68 conjectures) — Epoch: 2.941176 native-points
Not used in the score. Source category: independent. Review status: reviewed-independent.
Configuration: gpt-6-astra_max; task version
Metric: mean score percentage, FrontierMath-Erdos
Original source · Retrieved 2026-09-06
Exact Epoch task version; selected best mean score across reported reasoning configurations. No pooling of task versions or substitution of best scorer output.
Box Complex Work Eval (overall): 77%
Not used in the score. Source category not recorded. Review status: unresolved.
Not admitted after the September 6 model review: the collected 77 for GPT-6 Astra on Box Complex Work Eval (overall) still lacks a matched source entry establishing its checkpoint and benchmark-specific metric, tools and retry setup. This is a citation/comparability gap, not a finding that the result is false or useless. More or maximum reasoning effort is not required.
FrontierCyber (Sep 2026) — Irregular: 86 native-points
Not used in the score. Source category not recorded. Review status: unresolved.
Not admitted after the September 6 model review: the collected 86 for GPT-6 Astra on FrontierCyber (Sep 2026) — Irregular still lacks a matched source entry establishing its checkpoint and benchmark-specific metric, tools and retry setup. This is a citation/comparability gap, not a finding that the result is false or useless. More or maximum reasoning effort is not required.
CyScenarioBench (Sep 2026) — Irregular: 59%
Not used in the score. Source category not recorded. Review status: unresolved.
Not admitted after the September 6 model review: the collected 59 for GPT-6 Astra on CyScenarioBench (Sep 2026) — Irregular still lacks a matched source entry establishing its checkpoint and benchmark-specific metric, tools and retry setup. This is a citation/comparability gap, not a finding that the result is false or useless. More or maximum reasoning effort is not required.
BioMysteryBench v1 — Vals: 79.26%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: openai_gpt-6-astra; Vals max effort
Metric: biomysterybench overall percentage in Vals benchmark harness
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
Tax Agent Bench v1 — Vals: 63.34%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: openai_gpt-6-astra; Vals max effort
Metric: tax_agent_bench overall percentage in Vals benchmark harness
Original source · Retrieved 2026-09-06
Reconciled against the exact model/configuration in the evaluator’s public source. Updated source value recorded explicitly; best published reasoning mode selected within the same benchmark protocol.
RuneBench (30m, mean ln(1 + XP/min)): 7.258225 native-points
Contributes to the score. Source category: independent. Review status: reviewed-archive.
Configuration: gpt6astra-high
Metric: Arithmetic mean of ln(1 + peakXpRate) over all 16 skills, exactly as the owner’s Heatmap.js default ranking. Data rates are already normalized; do not divide by 200 again.
Original source · Retrieved 2026-09-06
Matched exact metric/value and the archived benchmark-specific configuration review. Unknown native effort stays explicitly unknown.
VoxelBench (text-to-voxel, Glicko-2): 2714 native-points
Contributes to the score. Source category: independent. Review status: reviewed-archive.
Configuration: GPT-6 Astra (Max)
Metric: Glicko-2 rating from the text-to-voxel board; not win percentage, deviation, image-input mode or MineBench Elo.
Original source · Retrieved 2026-09-06
Matched exact metric/value and the archived benchmark-specific configuration review. Unknown native effort stays explicitly unknown.
Roboflow Vision Evals (six-task mean, high tier): 88.533333%
Contributes to the score. Source category: independent. Review status: reviewed-archive.
Configuration: Require a complete six-task high-tier panel, the highest published benchmark tier. Retain high-tier scores even when lower than low-tier. Do not substitute the best of three repetitions: use the published mean. Each panel enters the fit once; do not also fit its components.
Metric: Arithmetic mean of six published high-tier task means; derived by The Latent, not Roboflow’s official low-tier overall score.
Original source · Retrieved 2026-09-06
Matched exact metric/value and the archived benchmark-specific configuration review. Unknown native effort stays explicitly unknown.
Code Arena WebDev (overall): 1797 native-points
Contributes to the score. Source category: independent. Review status: reviewed-archive.
Configuration: gpt-6-astra-max
Metric: Published overall WebDev preference rating, not task success percentage or Arena Text score.
Original source · Retrieved 2026-09-06
Matched exact metric/value and the archived benchmark-specific configuration review. Unknown native effort stays explicitly unknown.
AA-AnalystAgent: 51.25%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max); Artificial Analysis published evaluation harness
Metric: accuracy, AA
Original source · Retrieved 2026-09-10
Exact checkpoint, metric and published aggregate matched to the evaluator source. Date is retrieval date; first publication date is not inferred.
AA-LCR v1.1: 80.666667%
Not used in the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max)
Metric: lcr
Original source · Retrieved 2026-09-10
AA v4.3 evaluation, kept separate from prior benchmark versions and other agent harnesses.
Agent Arena overall IPS: 0.125451 native-points
Not used in the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (Max); Agent Arena 2026-09-08
Metric: IPS score (tau-hat), not task success percentage
Original source · Retrieved 2026-09-10
Native inverse-propensity-score metric. Retained as a separate uncalibrated benchmark, not converted to an accuracy percentage or Text Arena Elo.
Astra-26 LiveBrowseComp — native web: 42.307692%
Not used in the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra medium; Codex CLI 0.153.4; built-in web
Metric: correct answers / all 26 tasks
Original source · Retrieved 2026-09-10
Native web reference configuration only. Exa, Keenable, Parallel, and Valyu are different tool stacks, not interchangeable reasoning modes. One run per task; not the existing BrowseComp benchmark.
AutomationBench-AA: 68.491748%
Not used in the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max)
Metric: automationBenchPartialScore
Original source · Retrieved 2026-09-10
AA v4.3 evaluation, kept separate from prior benchmark versions and other agent harnesses.
Creative Writing v3: 2163.9 native-points
Contributes to the score. Source category: independent. Review status: reviewed-archive.
Configuration: gpt-6-astra; published evaluation, effort unspecified
Metric: Creative Writing v3 Elo
Original source · Retrieved 2026-09-10
Star denotes community-submitted evaluation hosted on the official benchmark board; not independently rerun by the benchmark author.
GDP.pdf — AA harness: 32.2%
Not used in the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (xhigh)
Metric: gdpPdfAllPass
Original source · Retrieved 2026-09-10
AA v4.3 evaluation, kept separate from prior benchmark versions and other agent harnesses.
LiveBench: 82.160643%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: gpt-6-astra-max; LiveBench 2026-06-25 task set
Metric: mean of seven category means
Original source · Retrieved 2026-09-10
Derived exactly using the existing category mapping and evaluator-published task scores; not a new benchmark version.
MathArena — ArXivLean Jun 2026: 63.83%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max); MathArena June 2026 competition harness
Metric: Accuracy
Original source · Retrieved 2026-09-10
Exact checkpoint, metric and published aggregate matched to the evaluator source. Date is retrieval date; first publication date is not inferred.
MathArena — ArXivMath Jun 2026: 94.44%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max); MathArena named monthly competition
Metric: Accuracy
Original source · Retrieved 2026-09-10
Exact published two-decimal accuracy from the direct model comparison. Month is kept explicit; overall values are not mixed with month-level rows.
MathArena — ArXivMath May 2026: 95%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max); MathArena named monthly competition
Metric: Accuracy
Original source · Retrieved 2026-09-10
Exact published two-decimal accuracy from the direct model comparison. Month is kept explicit; overall values are not mixed with month-level rows.
MathArena — BrokenArXiv Jun 2026: 99.07%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max); MathArena named monthly competition
Metric: Accuracy
Original source · Retrieved 2026-09-10
Exact published two-decimal accuracy from the direct model comparison. Month is kept explicit; overall values are not mixed with month-level rows.
MLCR-AA (Medical Long Context Reasoning): 35%
Contributes to the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (max); Artificial Analysis published evaluation harness
Metric: overall accuracy, AA
Original source · Retrieved 2026-09-10
Exact checkpoint, metric and published aggregate matched to the evaluator source. Date is retrieval date; first publication date is not inferred.
Terminal-Bench 4.0 — AA harness: 59.59596%
Not used in the score. Source category: independent. Review status: reviewed-independent.
Configuration: GPT-6 Astra (xhigh)
Metric: terminalbenchV40
Original source · Retrieved 2026-09-10
AA v4.3 evaluation, kept separate from prior benchmark versions and other agent harnesses.
Cite this model record
The Latent. “GPT-6 Astra: score and benchmark evidence.” 2026-09-10. Release tli-2026-v1.0-2026-09-10-benchmarks-1.
Download release data and definitions (JSON) · Data usage terms
Related reporting
These dated stories provide reporting context; they may describe an earlier release.