Compare · ModelsLive · 3 picked · head to head
Gemini 3.1 Pro Preview vs Mistral Medium 3.5 vs Qwen3.6 35B A3B
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Gemini 3.1 Pro Preview wins on 23/26 benchmarks
Gemini 3.1 Pro Preview wins 23 of 26 shared benchmarks. Leads in speed · general · coding.
Category leads
speed·Gemini 3.1 Pro Previewgeneral·Gemini 3.1 Pro Previewcoding·Gemini 3.1 Pro Previewarena·Gemini 3.1 Pro Previewknowledge·Gemini 3.1 Pro Previewmath·Gemini 3.1 Pro Preview
Hype vs Reality
Attention vs performance
Gemini 3.1 Pro Preview
#137 by perf·#5 by attention
Mistral Medium 3.5
#250 by perf·no signal
Qwen3.6 35B A3B
#178 by perf·#2 by attention
Best value
Qwen3.6 35B A3B
10.5x better value than Gemini 3.1 Pro Preview
Gemini 3.1 Pro Preview
6.9 pts/$
$7.00/M
Mistral Medium 3.5
6.6 pts/$
$4.50/M
Qwen3.6 35B A3B
72.3 pts/$
$0.57/M
Vendor risk
Who is behind the model
Google DeepMind
$4.20T·Tier 1
Mistral AI
$14.0B·Tier 1
Alibaba (Qwen)
$293.0B·Tier 1
Head to head
26 benchmarks · 3 models
Gemini 3.1 Pro PreviewMistral Medium 3.5Qwen3.6 35B A3B
Artificial Analysis · Agentic Index
Qwen3.6 35B A3B leads by +0.0
Artificial Analysis Agentic Index · a composite score measuring how well a model performs in agentic workflows · multi-step tool use, planning, error recovery, and autonomous task completion. Aggregates results from multiple agentic benchmarks including SWE-bench, tool-use tests, and planning evaluations. The canonical single-number metric for "how good is this model as an agent?"
Gemini 3.1 Pro Preview
21.4
Mistral Medium 3.5
19.0
Qwen3.6 35B A3B
21.4
Artificial Analysis · Coding Index
Gemini 3.1 Pro Preview leads by +21.9
Artificial Analysis Coding Index · a composite score that aggregates performance across multiple coding benchmarks into a single index. Tracks code generation quality, debugging ability, multi-language competence, and real-world software engineering tasks. Used by Artificial Analysis to rank model coding capability in a normalized, comparable format. Useful for developers choosing between models for coding-heavy workloads.
Gemini 3.1 Pro Preview
68.8
Mistral Medium 3.5
46.9
Qwen3.6 35B A3B
41.9
Artificial Analysis · CritPt
Gemini 3.1 Pro Preview leads by +17.4
Gemini 3.1 Pro Preview
17.7
Mistral Medium 3.5
0.0
Qwen3.6 35B A3B
0.3
Artificial Analysis · GDPval
Qwen3.6 35B A3B leads by +5.2
Gemini 3.1 Pro Preview
13.8
Mistral Medium 3.5
12.4
Qwen3.6 35B A3B
19.0
Artificial Analysis · GPQA Diamond
Gemini 3.1 Pro Preview leads by +10.0
Gemini 3.1 Pro Preview
94.1
Mistral Medium 3.5
74.8
Qwen3.6 35B A3B
84.1
Artificial Analysis · Humanity's Last Exam
Gemini 3.1 Pro Preview leads by +24.8
Gemini 3.1 Pro Preview
47.0
Mistral Medium 3.5
13.8
Qwen3.6 35B A3B
22.2
Artificial Analysis · IFBench
Gemini 3.1 Pro Preview leads by +8.3
Gemini 3.1 Pro Preview
77.1
Mistral Medium 3.5
68.8
Qwen3.6 35B A3B
64.4
Artificial Analysis · Long Context Reasoning
Gemini 3.1 Pro Preview leads by +10.3
Gemini 3.1 Pro Preview
82.0
Mistral Medium 3.5
69.3
Qwen3.6 35B A3B
71.7
Artificial Analysis · MMMU Pro
Gemini 3.1 Pro Preview leads by +7.4
Gemini 3.1 Pro Preview
82.4
Mistral Medium 3.5
64.9
Qwen3.6 35B A3B
75.0
Artificial Analysis · Quality Index
Gemini 3.1 Pro Preview leads by +11.5
Gemini 3.1 Pro Preview
29.7
Mistral Medium 3.5
14.2
Qwen3.6 35B A3B
18.2
Artificial Analysis · SciCode
Gemini 3.1 Pro Preview leads by +18.5
Gemini 3.1 Pro Preview
58.7
Mistral Medium 3.5
40.2
Qwen3.6 35B A3B
36.6
Artificial Analysis · tau2-Bench Telecom
Gemini 3.1 Pro Preview leads by +0.3
Gemini 3.1 Pro Preview
95.6
Mistral Medium 3.5
94.2
Qwen3.6 35B A3B
95.3
Artificial Analysis · Terminal-Bench Hard
Gemini 3.1 Pro Preview leads by +19.0
Gemini 3.1 Pro Preview
53.8
Mistral Medium 3.5
33.3
Qwen3.6 35B A3B
34.8
Dtbench
Gemini 3.1 Pro Preview leads by +36.0
Gemini 3.1 Pro Preview
95.1
Mistral Medium 3.5
59.1
Qwen3.6 35B A3B
56.5
Lmca
Gemini 3.1 Pro Preview leads by +28.4
Gemini 3.1 Pro Preview
63.3
Mistral Medium 3.5
30.7
Qwen3.6 35B A3B
34.9
WeirdML
Gemini 3.1 Pro Preview leads by +28.4
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Gemini 3.1 Pro Preview
72.1
Mistral Medium 3.5
43.7
Qwen3.6 35B A3B
34.5
Chatbot Arena Elo · Coding
Gemini 3.1 Pro Preview leads by +183.3
Gemini 3.1 Pro Preview
1446.2
Mistral Medium 3.5
1262.9
Chatbot Arena Elo · Overall
Gemini 3.1 Pro Preview leads by +60.0
Gemini 3.1 Pro Preview
1487.0
Mistral Medium 3.5
1426.9
Chess Puzzles
Gemini 3.1 Pro Preview leads by +30.5
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
Gemini 3.1 Pro Preview
52.6
Qwen3.6 35B A3B
22.1
FrontierMath-Tiers-1-3-v2-Private
Gemini 3.1 Pro Preview leads by +39.3
Gemini 3.1 Pro Preview
59.6
Qwen3.6 35B A3B
20.4
GPQA diamond
Gemini 3.1 Pro Preview leads by +12.8
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Gemini 3.1 Pro Preview
92.6
Qwen3.6 35B A3B
79.8
Mystery Game Puzzles
Gemini 3.1 Pro Preview leads by +13.2
Gemini 3.1 Pro Preview
27.3
Qwen3.6 35B A3B
14.1
OTIS Mock AIME 2024-2025
Gemini 3.1 Pro Preview leads by +8.9
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Gemini 3.1 Pro Preview
95.6
Qwen3.6 35B A3B
86.7
Proofbench
Gemini 3.1 Pro Preview leads by +17.0
Gemini 3.1 Pro Preview
26.0
Mistral Medium 3.5
9.0
Surface Evolver Bench
Qwen3.6 35B A3B leads by +17.5
Mistral Medium 3.5
26.9
Qwen3.6 35B A3B
44.4
Terminal Bench
Gemini 3.1 Pro Preview leads by +57.2
Terminal-Bench 2.0 · evaluates AI agents on real terminal-based coding tasks · writing scripts, debugging, running tests, and managing projects entirely through command-line interaction. Tests both code quality and terminal fluency. Claude Opus 4.7 scores 69.4%, demonstrating significant agentic terminal competence.
Gemini 3.1 Pro Preview
80.2
Qwen3.6 35B A3B
23.0
Full benchmark table
| Benchmark | Gemini 3.1 Pro Preview | Mistral Medium 3.5 | Qwen3.6 35B A3B |
|---|---|---|---|
Artificial Analysis · Agentic Index Artificial Analysis Agentic Index · a composite score measuring how well a model performs in agentic workflows · multi-step tool use, planning, error recovery, and autonomous task completion. Aggregates results from multiple agentic benchmarks including SWE-bench, tool-use tests, and planning evaluations. The canonical single-number metric for "how good is this model as an agent?" | 21.4 | 19.0 | 21.4 |
Artificial Analysis · Coding Index Artificial Analysis Coding Index · a composite score that aggregates performance across multiple coding benchmarks into a single index. Tracks code generation quality, debugging ability, multi-language competence, and real-world software engineering tasks. Used by Artificial Analysis to rank model coding capability in a normalized, comparable format. Useful for developers choosing between models for coding-heavy workloads. | 68.8 | 46.9 | 41.9 |
Artificial Analysis · CritPt | 17.7 | 0.0 | 0.3 |
Artificial Analysis · GDPval | 13.8 | 12.4 | 19.0 |
Artificial Analysis · GPQA Diamond | 94.1 | 74.8 | 84.1 |
Artificial Analysis · Humanity's Last Exam | 47.0 | 13.8 | 22.2 |
Artificial Analysis · IFBench | 77.1 | 68.8 | 64.4 |
Artificial Analysis · Long Context Reasoning | 82.0 | 69.3 | 71.7 |
Artificial Analysis · MMMU Pro | 82.4 | 64.9 | 75.0 |
Artificial Analysis · Quality Index | 29.7 | 14.2 | 18.2 |
Artificial Analysis · SciCode | 58.7 | 40.2 | 36.6 |
Artificial Analysis · tau2-Bench Telecom | 95.6 | 94.2 | 95.3 |
Artificial Analysis · Terminal-Bench Hard | 53.8 | 33.3 | 34.8 |
Dtbench | 95.1 | 59.1 | 56.5 |
Lmca | 63.3 | 30.7 | 34.9 |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 72.1 | 43.7 | 34.5 |
Chatbot Arena Elo · Coding | 1446.2 | 1262.9 | — |
Chatbot Arena Elo · Overall | 1487.0 | 1426.9 | — |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 52.6 | — | 22.1 |
FrontierMath-Tiers-1-3-v2-Private | 59.6 | — | 20.4 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 92.6 | — | 79.8 |
Mystery Game Puzzles | 27.3 | — | 14.1 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 95.6 | — | 86.7 |
Proofbench | 26.0 | 9.0 | — |
Surface Evolver Bench | — | 26.9 | 44.4 |
Terminal Bench Terminal-Bench 2.0 · evaluates AI agents on real terminal-based coding tasks · writing scripts, debugging, running tests, and managing projects entirely through command-line interaction. Tests both code quality and terminal fluency. Claude Opus 4.7 scores 69.4%, demonstrating significant agentic terminal competence. | 80.2 | — | 23.0 |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $2.00 | $12.00 | 1.0M tokens (~524 books) | $45.00 | |
| $1.50 | $7.50 | 262K tokens (~131 books) | $30.00 | |
| $0.15 | $1.00 | 262K tokens (~131 books) | $3.63 |