Compare · ModelsLive · 3 picked · head to head
GPT-5 vs Kimi K2 0711 vs Qwen3 Max
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
GPT-5 wins on 18/21 benchmarks
GPT-5 wins 18 of 21 shared benchmarks. Leads in knowledge · coding · general.
Category leads
knowledge·GPT-5coding·GPT-5general·GPT-5math·GPT-5language·GPT-5reasoning·Kimi K2 0711
Hype vs Reality
Attention vs performance
GPT-5
#100 by perf·#4 by attention
Kimi K2 0711
#72 by perf·#17 by attention
Qwen3 Max
#161 by perf·#2 by attention
Best value
Kimi K2 0711
2.0x better value than Qwen3 Max
GPT-5
9.4 pts/$
$5.63/M
Kimi K2 0711
39.1 pts/$
$1.43/M
Qwen3 Max
19.1 pts/$
$2.34/M
Vendor risk
Who is behind the model
OpenAI
$840.0B·Tier 1
Moonshot AI
$18.0B·Tier 1
Alibaba (Qwen)
$293.0B·Tier 1
Head to head
21 benchmarks · 3 models
GPT-5Kimi K2 0711Qwen3 Max
Fiction.LiveBench
GPT-5 leads by +30.5
Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination.
GPT-5
97.2
Kimi K2 0711
61.1
Qwen3 Max
66.7
Lech Mazur Writing
Qwen3 Max leads by +1.1
Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication.
GPT-5
86.0
Kimi K2 0711
85.6
Qwen3 Max
87.1
Aider polyglot
GPT-5 leads by +28.9
Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework.
GPT-5
88.0
Kimi K2 0711
59.1
Chess Puzzles
GPT-5 leads by +33.7
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
GPT-5
33.7
Qwen3 Max
0.0
Dtbench
GPT-5 leads by +14.2
GPT-5
84.5
Qwen3 Max
70.2
FrontierMath-Tiers-1-3-v2-Private
GPT-5 leads by +36.5
GPT-5
55.4
Qwen3 Max
18.9
GPQA diamond
GPT-5 leads by +18.1
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
GPT-5
81.6
Qwen3 Max
63.5
GSO-Bench
GPT-5 leads by +2.0
GSO-Bench · evaluates AI models on real-world open-source software engineering tasks, testing the ability to understand and resolve actual GitHub issues.
GPT-5
6.9
Kimi K2 0711
4.9
HELM · GPQA
GPT-5 leads by +13.9
GPT-5
79.1
Kimi K2 0711
65.2
HELM · IFEval
GPT-5 leads by +2.5
GPT-5
87.5
Kimi K2 0711
85.0
HELM · MMLU-Pro
GPT-5 leads by +4.4
GPT-5
86.3
Kimi K2 0711
81.9
HELM · Omni-MATH
Kimi K2 0711 leads by +0.7
GPT-5
64.7
Kimi K2 0711
65.4
HELM · WildBench
Kimi K2 0711 leads by +0.5
GPT-5
85.7
Kimi K2 0711
86.2
Lmca
GPT-5 leads by +13.8
GPT-5
47.0
Qwen3 Max
33.3
MATH level 5
GPT-5 leads by +1.0
MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics.
GPT-5
98.1
Qwen3 Max
97.1
Mystery Game Puzzles
GPT-5 leads by +15.2
GPT-5
15.2
Qwen3 Max
0.0
OTIS Mock AIME 2024-2025
GPT-5 leads by +18.1
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
GPT-5
91.4
Qwen3 Max
73.3
SimpleBench
GPT-5 leads by +36.5
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
GPT-5
48.0
Kimi K2 0711
11.6
SimpleQA Verified
GPT-5 leads by +1.4
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
GPT-5
50.1
Qwen3 Max
48.8
Terminal Bench
GPT-5 leads by +21.8
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.
GPT-5
49.6
Kimi K2 0711
27.8
WeirdML
GPT-5 leads by +21.3
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
GPT-5
60.7
Kimi K2 0711
39.4
Full benchmark table
| Benchmark | GPT-5 | Kimi K2 0711 | Qwen3 Max |
|---|---|---|---|
Fiction.LiveBench Fiction.LiveBench · a continuously updated benchmark using recently published fiction to test reading comprehension and reasoning, preventing data contamination. | 97.2 | 61.1 | 66.7 |
Lech Mazur Writing Lech Mazur Writing · evaluates creative writing ability, assessing prose quality, narrative coherence, and stylistic sophistication. | 86.0 | 85.6 | 87.1 |
Aider polyglot Aider Polyglot · measures how well AI models can edit code across multiple programming languages using the Aider coding assistant framework. | 88.0 | 59.1 | — |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 33.7 | — | 0.0 |
Dtbench | 84.5 | — | 70.2 |
FrontierMath-Tiers-1-3-v2-Private | 55.4 | — | 18.9 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 81.6 | — | 63.5 |
GSO-Bench GSO-Bench · evaluates AI models on real-world open-source software engineering tasks, testing the ability to understand and resolve actual GitHub issues. | 6.9 | 4.9 | — |
HELM · GPQA | 79.1 | 65.2 | — |
HELM · IFEval | 87.5 | 85.0 | — |
HELM · MMLU-Pro | 86.3 | 81.9 | — |
HELM · Omni-MATH | 64.7 | 65.4 | — |
HELM · WildBench | 85.7 | 86.2 | — |
Lmca | 47.0 | — | 33.3 |
MATH level 5 MATH Level 5 · the hardest tier of the MATH benchmark, featuring competition-level problems from AMC, AIME, and Olympiad-style mathematics. | 98.1 | — | 97.1 |
Mystery Game Puzzles | 15.2 | — | 0.0 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 91.4 | — | 73.3 |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 48.0 | 11.6 | — |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 50.1 | — | 48.8 |
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. | 49.6 | 27.8 | — |
WeirdML WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns. | 60.7 | 39.4 | — |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $1.25 | $10.00 | 400K tokens (~200 books) | $34.38 | |
| $0.57 | $2.30 | 131K tokens (~66 books) | $10.03 | |
| $0.78 | $3.90 | 262K tokens (~131 books) | $15.60 |