Gemini 3.8 Flash vs GPT-6 Astra
Humanity's Last Exam Humanity's Last Exam: Very hard expert questions across many fields, written to stump today’s models
Gemini 3.8 Flash42.1
GPT-6 Astra52.7
Terminal-Bench 4.0 Terminal-Bench 4.0 · source setting 2: Agents use a terminal to complete executable coding, system administration and research tasks in the 4.0 suite
Gemini 3.8 Flash19.7
GPT-6 Astra49.5
AutomationBench-AA AutomationBench-AA: Agents orchestrate SaaS workflows through APIs; objectives receive credit only when the task has no guardrail violation
Gemini 3.8 Flash60.9
GPT-6 Astra64.6
Arena Text · creative writing Arena Text · creative writing: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,487
GPT-6 Astra · max1,450
MMMU-Pro (vision) MMMU-Pro (vision): College-level questions that require understanding images: charts, diagrams, photos
Gemini 3.8 Flash84.2
GPT-6 Astra85.1
| Benchmark | Gemini 3.8 Flash | GPT-6 Astra |
|---|---|---|
| Humanity's Last Exam | 42.1 | 52.7 |
| Terminal-Bench 4.0 | 19.7 | 49.5 |
| AutomationBench-AA | 60.9 | 64.6 |
| Arena Text · creative writing | 1,487 · high | 1,450 · max |
| MMMU-Pro (vision) | 84.2 | 85.1 |
Specs
| Spec | Gemini 3.8 Flash | GPT-6 Astra |
|---|---|---|
| Input, per 1M | !$0.75 | $10 |
| Output, per 1M | !$3.75 | $50 |
| Cost per task? | $0.93 | $1.54 |
| Speed, tok/s | — | 59 |
| Context | 1M | 1.05M |
| Weights | Closed | Closed |
| Released | Sep 2026 | Sep 2026 |
| See images | ✓ | ✓ |
| Read PDFs | ✓ | Partial |
| Hear audio | ✓ | — |
| Watch video | ✓ | — |
| Call tools | ✓ | ✓ |
| Force a tool call | Partial | Partial |
| Return strict JSON | ✓ | ✓ |
| Search the web | ✓ | ✓ |
| Run code | ✓ | ✓ |
| Use a computer | ? | ✓ |
What breaks when you switch
Gemini 3.8 Flash → GPT-6 Astra
Monitoring can return HTTP 403›
Misalignment monitoring can return HTTP 403 before streaming or errors after output; match misalignment_policy_violation and handle stream errors.
No none reasoning effort›
The none reasoning effort is unsupported on GPT-6 Astra and GPT-6.1 Sol; use low instead.
Remove unsupported sampling parameters›
When reasoning effort is not none, remove temperature, top_p and top_logprobs, plus Chat Completions logprobs or Responses output logprobs inclusion.
GPT-6 Astra → Gemini 3.8 Flash
Context: 1.05M → 1M
Sampling parameters are ignored›
temperature, top_p and top_k are ignored now and will cause HTTP 400 in future generations; remove them from requests.
No minimal thinking level›
The minimal thinking level returns an error; avoid minimal.