Gemini 3.8 Flash vs GPT-6.1 Sol
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-6.1 Sol49.9
Arena Text · overall Arena Text · overall: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,496
GPT-6.1 Sol · max1,484
Arena Text · Chinese Arena Text · chinese: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,550.8
GPT-6.1 Sol · max1,494.6
Arena Text · English Arena Text · english: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,499
GPT-6.1 Sol · max1,454.2
Arena Text · French Arena Text · french: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,491.7
GPT-6.1 Sol · max1,470.1
Arena Text · instruction following Arena Text · instruction following: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,491
GPT-6.1 Sol · max1,468.6
Arena Text · non-English Arena Text · non english: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,491.1
GPT-6.1 Sol · max1,437.5
Arena Text · Russian Arena Text · russian: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,514.4
GPT-6.1 Sol · max1,457.9
Arena Text · Spanish Arena Text · spanish: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,485.7
GPT-6.1 Sol · max1,440.4
AA Omniscience Index AA Omniscience Index: Scores factual knowledge while penalizing incorrect answers to encourage admitting uncertainty
Gemini 3.8 Flash28.6
GPT-6.1 Sol40
SimpleQA Verified v1 SimpleQA Verified v1: Accuracy on short factual questions with verified answers, answered without web access
Gemini 3.8 Flash74.6
GPT-6.1 Sol · max73.9
CritPt (physics) CritPt (physics): Research-level physics problems that need many careful reasoning steps
Gemini 3.8 Flash12.3
GPT-6.1 Sol27.7
AA Intelligence Index 4.3.2 AA Intelligence Index 4.3.2: Artificial Analysis combines tests of knowledge, reasoning, coding and tool use into one score
Gemini 3.8 Flash39.8
GPT-6.1 Sol47.8
AA Omniscience · Accuracy AA Omniscience · Accuracy: Measures correct answers on the AA Omniscience factual knowledge test
Gemini 3.8 Flash53
GPT-6.1 Sol60.4
AA Omniscience · Hallucination Rate AA Omniscience · Hallucination Rate: Measures how often attempted answers are incorrect on the AA Omniscience factual knowledge test · lower is better
Gemini 3.8 Flash51.9
GPT-6.1 Sol51.6
AA-LCR (long context) 1.1 AA-LCR (long context) 1.1: Questions that need reading and connecting facts across very long documents (around 100K tokens)
Gemini 3.8 Flash84
GPT-6.1 Sol83.3
GDP.pdf · AA all-pass GDP.pdf · AA all-pass: Models reason over professional PDF documents; an attempt passes only when every atomic criterion passes
Gemini 3.8 Flash22.8
GPT-6.1 Sol30
MLCR-AA MLCR-AA: Models synthesize long medical records; the headline credits concise answers judged both accurate and complete
Gemini 3.8 Flash · high21.7
GPT-6.1 Sol · max33.9
ARC-AGI-1 · public evaluation ARC-AGI-1 · public-eval: Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit
Gemini 3.8 Flash97.8
GPT-6.1 Sol97.9
ARC-AGI-1 · semi-private ARC-AGI-1 · semi-private: Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit
Gemini 3.8 Flash97.5
GPT-6.1 Sol95.5
ARC-AGI-2 · public evaluation ARC-AGI-2 · public-eval: Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit
Gemini 3.8 Flash87.5
GPT-6.1 Sol90.6
ARC-AGI-2 · semi-private ARC-AGI-2 · semi-private: Models infer transformations from examples and solve unseen grid puzzles; the evaluation split is explicit
Gemini 3.8 Flash82.9
GPT-6.1 Sol86.7
ARC-AGI-3 · semi-private ARC-AGI-3 · semi-private: Measures interactive game solving through relative human action efficiency
Gemini 3.8 Flash4
GPT-6.1 Sol10.6
Chess Puzzles Chess Puzzles: Models choose the best next move from a novel chess position
Gemini 3.8 Flash · high58
GPT-6.1 Sol · max68
FrontierMath Erdos FrontierMath Erdos: Models solve expert-written advanced mathematics problems on the named tier and edition
Gemini 3.8 Flash · high0
GPT-6.1 Sol · max4.4
FrontierMath Tier 4 (v2) FrontierMath Tier 4 (v2): Models solve expert-written advanced mathematics problems on the named tier and edition
Gemini 3.8 Flash · high22
GPT-6.1 Sol · max100
FrontierMath Tiers 1–3 (v2) FrontierMath Tiers 1–3 (v2): Models solve expert-written advanced mathematics problems on the named tier and edition
Gemini 3.8 Flash · high68.4
GPT-6.1 Sol · max93.7
Mystery Game Puzzles Mystery Game Puzzles: Models choose the best move in puzzle positions from an undisclosed game
Gemini 3.8 Flash · high47
GPT-6.1 Sol · max80
MathArena · ArXivMath · August 2026 MathArena · arxiv · 08/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3.8 Flash40.4
GPT-6.1 Sol · max96.5
MathArena · ArXivMath · June 2026 MathArena · arxiv · 06/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3.8 Flash67.4
GPT-6.1 Sol · max94.4
MathArena · ArXivMath · May 2026 MathArena · arxiv · 05/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3.8 Flash58.3
GPT-6.1 Sol · max95
MathArena · ArXivMath · September 2026 MathArena · arxiv · 09/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3.8 Flash50
GPT-6.1 Sol · max91.5
MathArena · BrokenArXiv · August 2026 MathArena · arxiv_false · 08/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3.8 Flash27.4
GPT-6.1 Sol · max86.3
MathArena · BrokenArXiv · June 2026 MathArena · arxiv_false · 06/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3.8 Flash20.8
GPT-6.1 Sol · max100
MathArena · BrokenArXiv · May 2026 MathArena · arxiv_false · 05/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3.8 Flash19
GPT-6.1 Sol · max90
MathArena · BrokenArXiv · September 2026 MathArena · arxiv_false · 09/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3.8 Flash26.7
GPT-6.1 Sol · max82.7
Vals MedCode Vals MedCode: Medical coding accuracy on patient encounters
Gemini 3.8 Flash48.1
GPT-6.1 Sol48.8
Vals MedScribe Vals MedScribe: Accuracy of clinical notes produced from consultation transcripts
Gemini 3.8 Flash84.5
GPT-6.1 Sol86.5
Vals Legal Research Bench Vals Legal Research Bench: Grounded legal research; a task passes only if all criteria pass
Gemini 3.8 Flash38.9
GPT-6.1 Sol38.5
Vals Science Index Vals Science Index: Scientific problem solving across the index’s component evaluations
Gemini 3.8 Flash35.8
GPT-6.1 Sol59.6
Arena Text · German Arena Text · german: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,497.6
GPT-6.1 Sol—
Arena Text · Japanese Arena Text · japanese: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,506.3
GPT-6.1 Sol—
Arena Text · Korean Arena Text · korean: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,461.6
GPT-6.1 Sol—
Arena Text · Polish Arena Text · polish: Human pairwise preferences in the stated text vote pool
Gemini 3.8 Flash · high1,498.5
GPT-6.1 Sol—
MathArena · ArXivMath · April 2026 MathArena · arxiv · 04/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3.8 Flash64.2
GPT-6.1 Sol—
MathArena · BrokenArXiv · April 2026 MathArena · arxiv_false · 04/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3.8 Flash33.6
GPT-6.1 Sol—
MathArena · ArXivLean · June 2026 MathArena · arxivlean · 06/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3.8 Flash19.6
GPT-6.1 Sol—
MathArena · ArXivLean · September 2026 MathArena · arxivlean · 09/2026: Answer accuracy on the named mathematics competition under the recorded effort
Gemini 3.8 Flash—
GPT-6.1 Sol · max35.9
Vals LegalBench Vals LegalBench: Legal reasoning and legal-language understanding on LegalBench
Gemini 3.8 Flash87
GPT-6.1 Sol—
| Benchmark | Gemini 3.8 Flash | GPT-6.1 Sol |
|---|---|---|
| Humanity's Last Exam | 42.1 | 49.9 |
| Arena Text · overall | 1,496 · high | 1,484 · max |
| Arena Text · Chinese | 1,550.8 · high | 1,494.6 · max |
| Arena Text · English | 1,499 · high | 1,454.2 · max |
| Arena Text · French | 1,491.7 · high | 1,470.1 · max |
| Arena Text · instruction following | 1,491 · high | 1,468.6 · max |
| Arena Text · non-English | 1,491.1 · high | 1,437.5 · max |
| Arena Text · Russian | 1,514.4 · high | 1,457.9 · max |
| Arena Text · Spanish | 1,485.7 · high | 1,440.4 · max |
| AA Omniscience Index | 28.6 | 40 |
| SimpleQA Verified v1 | 74.6 | 73.9 · max |
| CritPt (physics) | 12.3 | 27.7 |
| AA Intelligence Index 4.3.2 | 39.8 | 47.8 |
| AA Omniscience · Accuracy | 53 | 60.4 |
| AA Omniscience · Hallucination Rate | 51.9 | 51.6 |
| AA-LCR (long context) 1.1 | 84 | 83.3 |
| GDP.pdf · AA all-pass | 22.8 | 30 |
| MLCR-AA | 21.7 · high | 33.9 · max |
| ARC-AGI-1 · public evaluation | 97.8 | 97.9 |
| ARC-AGI-1 · semi-private | 97.5 | 95.5 |
| ARC-AGI-2 · public evaluation | 87.5 | 90.6 |
| ARC-AGI-2 · semi-private | 82.9 | 86.7 |
| ARC-AGI-3 · semi-private | 4 | 10.6 |
| Chess Puzzles | 58 · high | 68 · max |
| FrontierMath Erdos | 0 · high | 4.4 · max |
| FrontierMath Tier 4 (v2) | 22 · high | 100 · max |
| FrontierMath Tiers 1–3 (v2) | 68.4 · high | 93.7 · max |
| Mystery Game Puzzles | 47 · high | 80 · max |
| MathArena · ArXivMath · August 2026 | 40.4 | 96.5 · max |
| MathArena · ArXivMath · June 2026 | 67.4 | 94.4 · max |
| MathArena · ArXivMath · May 2026 | 58.3 | 95 · max |
| MathArena · ArXivMath · September 2026 | 50 | 91.5 · max |
| MathArena · BrokenArXiv · August 2026 | 27.4 | 86.3 · max |
| MathArena · BrokenArXiv · June 2026 | 20.8 | 100 · max |
| MathArena · BrokenArXiv · May 2026 | 19 | 90 · max |
| MathArena · BrokenArXiv · September 2026 | 26.7 | 82.7 · max |
| Vals MedCode | 48.1 | 48.8 |
| Vals MedScribe | 84.5 | 86.5 |
| Vals Legal Research Bench | 38.9 | 38.5 |
| Vals Science Index | 35.8 | 59.6 |
| Arena Text · German | 1,497.6 · high | — |
| Arena Text · Japanese | 1,506.3 · high | — |
| Arena Text · Korean | 1,461.6 · high | — |
| Arena Text · Polish | 1,498.5 · high | — |
| MathArena · ArXivMath · April 2026 | 64.2 | — |
| MathArena · BrokenArXiv · April 2026 | 33.6 | — |
| MathArena · ArXivLean · June 2026 | 19.6 | — |
| MathArena · ArXivLean · September 2026 | — | 35.9 · max |
| Vals LegalBench | 87 | — |
Specs
| Spec | Gemini 3.8 Flash | GPT-6.1 Sol |
|---|---|---|
| Input, per 1M | !$0.75 | $2 |
| Output, per 1M | !$3.75 | $10 |
| Cost per task? | $0.93 | $0.21 |
| Speed, tok/s | — | 45 |
| 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-6.1 Sol
Remove unsupported parameters›
With reasoning enabled, remove temperature, top_p and top_logprobs, plus Chat Completions logprobs or Responses message.output_text.logprobs includes.
Responses API for tool calling · Chat Completions›
Use Responses for tools; Chat Completions supports this model without tool calling.
No none or minimal reasoning›
The none and minimal reasoning efforts are unsupported; choose low, medium, high, xhigh or max.
More than 272K input tokens›
Prompts above 272K input tokens charge 2× input and cache rates and 1.5× output rates for the entire request.
Cache writes billed at 1.25x›
Cache writes cost 1.25× the uncached input token rate.
Ultrafast prices 6x Standard›
Ultrafast mode costs 6× Standard token prices.
GPT-6.1 Sol → 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.