none reviewed yet · 2 benchmark observations · latest capture Sep 8, 2026
Inspect the evidenceOpenAI · MODEL PROFILE
GPT-6 Astra
Teams that want the single strongest one-shot frontend generator and can absorb frontier pricing
Saved editorial fit for your current task lens; not a benchmark result
How scoring worksVerified rates (Sep 8, 2026) · 100 runs / month at your token settings
Adjust workloadAt a glance
9 facts- Provider
- OpenAI
- Type
- API model
- Access
- OpenAI API (Chat Completions, Responses, Batch), Microsoft Azure, AWS Bedrock, ChatGPT Plus/Pro/Business/Enterprise; Codex harness
- Context
- 1,050,000 tokens
- Maximum output
- 128,000 tokens
- License / weights
- Proprietary
- Input modalities
- file, image, text
- Output modalities
- text
- Pricing notes
- $10 in / $50 out per 1M; cached input $1.00, cache write $12.50; above 272K context $20 in / $75 out; Fast mode 2x price; Batch $5/$25
The practical trade-offs
Documented strengths
- Top of Arena WebDev Frontend by a clear margin (1808)
- Best published Terminal-Bench 4.0 (57.9%) and long-context retrieval (MRCR v2 96.3% at 512K-1M)
- Strongest screen/UI grounding evidence: ScreenSpot-Pro 92.7%, OSWorld 2.0 72.6%
- Lowest published hallucination rate in its cohort (4.2%)
Limitations to check
- Most expensive model here ($10/$50, doubling above 272K context)
- Very high latency at max effort (AA TTFT 317 s)
- No SWE-bench Pro/Verified numbers published
- Arena rank rests on only 1,055 votes (+/-26), so rank 1 vs 2 is inside the interval
Follow the source
7 source linksEvidence & conditions
7Source links
NoneReviewed records
2Benchmark runs
Sep 8, 2026Latest capture
- Evidence source 1
developers.openai.comDate not recorded - Evidence source 2
openai.comDate not recorded - Evidence source 3
openai.comDate not recorded - Evidence source 4
arena.aiDate not recorded - Evidence source 5
artificialanalysis.aiDate not recorded - Evidence source 6
computingforgeeks.comDate not recorded - Evidence source 7
openrouter.aiDate not recorded
Benchmark observations
Terminal-Bench 4.0 Inspect conditions
- name
- Terminal-Bench 4.0
- value
- 57.9
- unit
- %
- source Url
- https://openai.com/index/gpt-6-astra/
- checked At
- 2026-09-08T19:30:56Z
- attribution
- provider-reported
- configuration
- OpenAI release table; maximum across effort; research/API evaluation, not production ChatGPT.
DeepSWE v1.1 Inspect conditions
- name
- DeepSWE v1.1
- value
- 74.1
- unit
- %
- source Url
- https://openai.com/index/gpt-6-astra/
- checked At
- 2026-09-08T19:30:56Z
- attribution
- provider-reported
- configuration
- OpenAI release table; maximum across effort; research/API evaluation, not production ChatGPT.
Source reports, measured results, and editorial assessments are separate evidence types. Source dates do not imply a new benchmark run. Capture dates are shown when recorded; otherwise we say the capture date was not recorded. Recorded dates are compared with this catalogue snapshot (Sep 11, 2026), not today’s date.
Method AIS-V2
How this page is assessed
Quality claims follow Method AIS-V2.
Money never buys rank — commercial relationships cannot change rank, eligibility, evidence quality, or publication timing.
Submit or claim a listing is free and does not skip the editorial queue.
Private · this device only