GPT-6 Astra

Maximum reported across reasoning efforts; OpenAI research environment or API

Provider-published result; comparator measurements are not automatically independently reproduced.

GPT-6 Astra: A new generation of intelligence · Professional table / Internal Data Science Tasks / GPT‑6 Astra · reviewed 2026-09-06

Maximum reported across reasoning efforts; OpenAI research environment or API

Provider-published result; comparator measurements are not automatically independently reproduced.

GPT-6 Astra: A new generation of intelligence · Coding table / Internal Database Migration Tasks / GPT‑6 Astra · reviewed 2026-09-06

Maximum reported across reasoning efforts; OpenAI research environment or API

Provider-published result; comparator measurements are not automatically independently reproduced.

GPT-6 Astra: A new generation of intelligence · Professional table / Internal Design Tasks / GPT‑6 Astra · reviewed 2026-09-06

System-card reported configuration; reasoning effort unspecified

Provider-published result; preserves source setting without claiming a controlled cross-provider comparison.

GPT-6 Astra System Card · Section10.1.3.1 · reviewed 2026-09-06

Comparator maximum available thinking/reasoning when reported, otherwise best available result; exact setting not identified by this grid. All models self-computed with Linux terminal, pre-installed bioinformatics tools, Python, R and internet access.

Provider-published launch claim, not independent measurement by Google for every row. All models self-computed with Linux terminal, pre-installed bioinformatics tools, Python, R and internet access. Methodology https://deepmind.google/models/evals-methodology/gemini-4-argon, SHA256 ff1df6bdeddc4c0f48840c09a8a7813b10ae7ad9e053892da68dabf64f09bff7. Preserve source rounding. Methodology says capabilities as of September 2026 and results as of October 2026; review date is source retrieval, not evaluation date.

Comparison limit: Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison.

Google Gemini 4 Argon launch and evaluation methodology · Launch grid row 11: LABBench 2 / GPT-6 Astra; methodology Additional Details · reviewed 2026-09-30

Maximum reported across reasoning efforts; OpenAI research environment or API

Provider-published result; comparator measurements are not automatically independently reproduced.

GPT-6 Astra: A new generation of intelligence · Science And Health table / LifeSciBench / GPT‑6 Astra · reviewed 2026-09-06

Comparator maximum available thinking/reasoning when reported, otherwise best available result; exact setting not identified by this grid. All models self-computed without tools. Gemini 1 FPS; Astra 800 frames, Fable 300 frames and Opus 600 frames due to API limits. Unequal frame budgets are excluded from matched comparison.

Provider-published launch claim, not independent measurement by Google for every row. All models self-computed without tools. Gemini 1 FPS; Astra 800 frames, Fable 300 frames and Opus 600 frames due to API limits. Unequal frame budgets are excluded from matched comparison. Methodology https://deepmind.google/models/evals-methodology/gemini-4-argon, SHA256 ff1df6bdeddc4c0f48840c09a8a7813b10ae7ad9e053892da68dabf64f09bff7. Preserve source rounding. Methodology says capabilities as of September 2026 and results as of October 2026; review date is source retrieval, not evaluation date.

Comparison limit: Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison.

Google Gemini 4 Argon launch and evaluation methodology · Launch grid row 18: LVBench / GPT-6 Astra; methodology Additional Details · reviewed 2026-09-30

Maximum reported across reasoning efforts; OpenAI research environment or API

Provider-published result; comparator measurements are not automatically independently reproduced.

GPT-6 Astra: A new generation of intelligence · Science And Health table / MedChemBench (Internal) / GPT‑6 Astra · reviewed 2026-09-06

UK AISI; single forward pass; no chain of thought

Time-horizon estimate; possible contamination noted by evaluator. Higher means harder tasks solved, not slower inference.

GPT-6 Astra System Card · Section9.3 · reviewed 2026-09-06

Maximum reported across reasoning efforts; OpenAI research environment or API; 8-needle 256K-512K

Provider-published result; comparator measurements are not automatically independently reproduced.

GPT-6 Astra: A new generation of intelligence · Long Context table / OpenAI MRCR v2 8-needle 256K-512K / GPT‑6 Astra · reviewed 2026-09-06

Maximum reported across reasoning efforts; OpenAI research environment or API; 8-needle 512K-1M

Provider-published result; comparator measurements are not automatically independently reproduced.

GPT-6 Astra: A new generation of intelligence · Long Context table / OpenAI MRCR v2 8-needle 512K-1M / GPT‑6 Astra · reviewed 2026-09-06

Maximum reported across reasoning efforts; OpenAI research environment or API

Provider-published result; comparator measurements are not automatically independently reproduced.

GPT-6 Astra: A new generation of intelligence · Professional table / OpenScore String Quartets (1 - OMR-NED) / GPT‑6 Astra · reviewed 2026-09-06

Comparator maximum available thinking/reasoning when reported, otherwise best available result; exact setting not identified by this grid. Argon self-computed, maximum of three single-attempt runs; offline subset partial score. Official Docker/evaluator, 1080p, 500 steps, Gemini CUA harness, parallel batch tools, compaction, pyautogui actuation, screenshot-only observations, UI-specific function declarations, safety filters, official 08.08 patch. Astra from its official blog. Anthropic combined online/offline results are excluded.

Provider-published launch claim, not independent measurement by Google for every row. Argon self-computed, maximum of three single-attempt runs; offline subset partial score. Official Docker/evaluator, 1080p, 500 steps, Gemini CUA harness, parallel batch tools, compaction, pyautogui actuation, screenshot-only observations, UI-specific function declarations, safety filters, official 08.08 patch. Astra from its official blog. Anthropic combined online/offline results are excluded. Methodology https://deepmind.google/models/evals-methodology/gemini-4-argon, SHA256 ff1df6bdeddc4c0f48840c09a8a7813b10ae7ad9e053892da68dabf64f09bff7. Preserve source rounding. Methodology says capabilities as of September 2026 and results as of October 2026; review date is source retrieval, not evaluation date.

Comparison limit: Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison.

Google Gemini 4 Argon launch and evaluation methodology · Launch grid row 16: OSWorld 2.0 offline partial score / GPT-6 Astra; methodology Additional Details · reviewed 2026-09-30

Maximum reported across reasoning efforts; OpenAI research environment or API; offline set; partial credit; v2026.08.08; official task/grading settings

Provider-published result; comparator measurements are not automatically independently reproduced.

GPT-6 Astra: A new generation of intelligence · Computer Use table / OSWorld 2.0 (v2026.08.08, offline set, partial score) / GPT‑6 Astra · reviewed 2026-09-06

Reported reasoning effort low. Offline set from v2026.08.08; partial reward, not binary full-task success or OSWorld Verified. OpenAI research environment or API; same named metric and evaluation chart. No fallback is reported for this OpenAI configuration.

Provider-published lab self-report. Published chart cost per task $2.7159; source-specific cost does not establish consensus workload efficiency. Competitor results are republished from public reports; the page does not establish independent reproduction by OpenAI.

Introducing GPT-6.1 Sol · Chart: OSWorld 2.0, offline set / GPT-6 Astra / low · reviewed 2026-09-29

Reported reasoning effort medium. Offline set from v2026.08.08; partial reward, not binary full-task success or OSWorld Verified. OpenAI research environment or API; same named metric and evaluation chart. No fallback is reported for this OpenAI configuration.

Provider-published lab self-report. Published chart cost per task $5.3594; source-specific cost does not establish consensus workload efficiency. Competitor results are republished from public reports; the page does not establish independent reproduction by OpenAI.

Introducing GPT-6.1 Sol · Chart: OSWorld 2.0, offline set / GPT-6 Astra / medium · reviewed 2026-09-29

Reported reasoning effort high. Offline set from v2026.08.08; partial reward, not binary full-task success or OSWorld Verified. OpenAI research environment or API; same named metric and evaluation chart. No fallback is reported for this OpenAI configuration.

Provider-published lab self-report. Published chart cost per task $6.9064; source-specific cost does not establish consensus workload efficiency. Competitor results are republished from public reports; the page does not establish independent reproduction by OpenAI.

Introducing GPT-6.1 Sol · Chart: OSWorld 2.0, offline set / GPT-6 Astra / high · reviewed 2026-09-29

Reported reasoning effort xhigh. Offline set from v2026.08.08; partial reward, not binary full-task success or OSWorld Verified. OpenAI research environment or API; same named metric and evaluation chart. No fallback is reported for this OpenAI configuration.

Provider-published lab self-report. Published chart cost per task $7.4944; source-specific cost does not establish consensus workload efficiency. Competitor results are republished from public reports; the page does not establish independent reproduction by OpenAI.

Introducing GPT-6.1 Sol · Chart: OSWorld 2.0, offline set / GPT-6 Astra / xhigh · reviewed 2026-09-29

Reported reasoning effort max. Offline set from v2026.08.08; partial reward, not binary full-task success or OSWorld Verified. OpenAI research environment or API; same named metric and evaluation chart. No fallback is reported for this OpenAI configuration.

Provider-published lab self-report. Published chart cost per task $9.4353; source-specific cost does not establish consensus workload efficiency. Competitor results are republished from public reports; the page does not establish independent reproduction by OpenAI.

Introducing GPT-6.1 Sol · Chart: OSWorld 2.0, offline set / GPT-6 Astra / max · reviewed 2026-09-29

System-card reported configuration; reasoning effort unspecified

Lower is better. Helpful-only checkpoint omitted as distinct noncatalog identity.

GPT-6 Astra System Card · Section10.1.1.2.4 · reviewed 2026-09-06

Comparator maximum available thinking/reasoning when reported, otherwise best available result; exact setting not identified by this grid. All models self-computed; OpenCode harness; 10-hour budget on one NVIDIA H100 GPU. Weighted aggregate across four base models and seven benchmarks.

Provider-published launch claim, not independent measurement by Google for every row. All models self-computed; OpenCode harness; 10-hour budget on one NVIDIA H100 GPU. Weighted aggregate across four base models and seven benchmarks. Methodology https://deepmind.google/models/evals-methodology/gemini-4-argon, SHA256 ff1df6bdeddc4c0f48840c09a8a7813b10ae7ad9e053892da68dabf64f09bff7. Preserve source rounding. Methodology says capabilities as of September 2026 and results as of October 2026; review date is source retrieval, not evaluation date.

Comparison limit: Provider-published claim. Independent source results are admitted separately with exact checkpoint, effort and compatible protocol; this grid does not establish a matched comparison.

Google Gemini 4 Argon launch and evaluation methodology · Launch grid row 9: PostTrainBench 1.1 / GPT-6 Astra; methodology Additional Details · reviewed 2026-09-30

observed performance

Provider-published result; preserves source setting without claiming a controlled cross-provider comparison.

GPT-6 Astra System Card · Section10.1.1.1 / ProtocolQA Open-Ended · reviewed 2026-09-06

refusal-adjusted upper estimate; refusals counted as successes

Provider-published result; preserves source setting without claiming a controlled cross-provider comparison.

GPT-6 Astra System Card · Section10.1.1.1 / ProtocolQA Open-Ended · reviewed 2026-09-06