Compare · ModelsLive · 3 picked · head to head
GLM 4.7 vs Kimi K2.5 vs Qwen3.5 397B A17B
Side by side · benchmarks, pricing, and signals you can act on.
Winner summary
Kimi K2.5 wins on 13/23 benchmarks
Kimi K2.5 wins 13 of 23 shared benchmarks. Leads in knowledge · math · language.
Category leads
agentic·Qwen3.5 397B A17Bknowledge·Kimi K2.5math·Kimi K2.5language·Kimi K2.5coding·GLM 4.7speed·Kimi K2.5arena·GLM 4.7general·Kimi K2.5reasoning·GLM 4.7
Hype vs Reality
Attention vs performance
GLM 4.7
#152 by perf·#3 by attention
Kimi K2.5
#122 by perf·#17 by attention
Qwen3.5 397B A17B
#48 by perf·#2 by attention
Best value
Kimi K2.5
1.1x better value than GLM 4.7
GLM 4.7
32.9 pts/$
$1.40/M
Kimi K2.5
37.0 pts/$
$1.35/M
Qwen3.5 397B A17B
29.6 pts/$
$2.02/M
Vendor risk
Who is behind the model
z-ai
private · undisclosed
Moonshot AI
$18.0B·Tier 1
Alibaba (Qwen)
$293.0B·Tier 1
Head to head
23 benchmarks · 3 models
GLM 4.7Kimi K2.5Qwen3.5 397B A17B
APEX-Agents
Qwen3.5 397B A17B leads by +10.5
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments.
GLM 4.7
3.1
Kimi K2.5
14.4
Qwen3.5 397B A17B
24.9
Chess Puzzles
Qwen3.5 397B A17B leads by +1.1
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
GLM 4.7
1.1
Kimi K2.5
7.4
Qwen3.5 397B A17B
8.5
GPQA diamond
Kimi K2.5 leads by +1.7
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
GLM 4.7
77.8
Kimi K2.5
83.5
Qwen3.5 397B A17B
81.8
OpenCompass · AIME2025
GLM 4.7 leads by +3.1
GLM 4.7
95.4
Kimi K2.5
91.9
Qwen3.5 397B A17B
92.3
OpenCompass · GPQA-Diamond
Qwen3.5 397B A17B leads by +0.3
GLM 4.7
86.9
Kimi K2.5
88.1
Qwen3.5 397B A17B
88.4
OpenCompass · HLE
Kimi K2.5 leads by +1.1
GLM 4.7
25.4
Kimi K2.5
28.6
Qwen3.5 397B A17B
27.5
OpenCompass · IFEval
Kimi K2.5 leads by +2.4
GLM 4.7
90.2
Kimi K2.5
93.9
Qwen3.5 397B A17B
91.5
OpenCompass · LiveCodeBenchV6
GLM 4.7 leads by +0.8
GLM 4.7
83.8
Kimi K2.5
80.6
Qwen3.5 397B A17B
83.0
OpenCompass · MMLU-Pro
Qwen3.5 397B A17B leads by +1.4
GLM 4.7
84.0
Kimi K2.5
86.2
Qwen3.5 397B A17B
87.6
OTIS Mock AIME 2024-2025
Kimi K2.5 leads by +3.3
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
GLM 4.7
83.3
Kimi K2.5
92.2
Qwen3.5 397B A17B
88.9
Artificial Analysis · Agentic Index
Kimi K2.5 leads by +39.1
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?"
Kimi K2.5
58.9
Qwen3.5 397B A17B
19.9
Artificial Analysis · Coding Index
Qwen3.5 397B A17B leads by +8.7
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.
Kimi K2.5
39.5
Qwen3.5 397B A17B
48.2
Artificial Analysis · Quality Index
Kimi K2.5 leads by +25.4
Kimi K2.5
46.8
Qwen3.5 397B A17B
21.4
Chatbot Arena Elo · Coding
GLM 4.7 leads by +35.0
GLM 4.7
1434.5
Qwen3.5 397B A17B
1399.4
Chatbot Arena Elo · Overall
Qwen3.5 397B A17B leads by +0.3
GLM 4.7
1441.5
Qwen3.5 397B A17B
1441.8
Cl Bench
Kimi K2.5 leads by +3.4
GLM 4.7
15.9
Kimi K2.5
19.3
Cl Bench Life
Kimi K2.5 leads by +2.3
GLM 4.7
10.9
Kimi K2.5
13.2
FrontierMath-2025-02-28-Private
Kimi K2.5 leads by +44.7
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
GLM 4.7
4.3
Kimi K2.5
49.0
FrontierMath-Tier-4-2025-07-01-Private
Kimi K2.5 leads by +7.0
FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning.
GLM 4.7
0.0
Kimi K2.5
7.0
PostTrainBench
Kimi K2.5 leads by +2.8
GLM 4.7
7.5
Kimi K2.5
10.3
SimpleBench
GLM 4.7 leads by +1.1
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
GLM 4.7
37.2
Kimi K2.5
36.2
SimpleQA Verified
Kimi K2.5 leads by +2.1
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
GLM 4.7
32.2
Kimi K2.5
34.3
Terminal Bench
Kimi K2.5 leads by +9.8
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.
GLM 4.7
33.4
Kimi K2.5
43.2
Full benchmark table
| Benchmark | GLM 4.7 | Kimi K2.5 | Qwen3.5 397B A17B |
|---|---|---|---|
APEX-Agents APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments. | 3.1 | 14.4 | 24.9 |
Chess Puzzles Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities. | 1.1 | 7.4 | 8.5 |
GPQA diamond Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs. | 77.8 | 83.5 | 81.8 |
OpenCompass · AIME2025 | 95.4 | 91.9 | 92.3 |
OpenCompass · GPQA-Diamond | 86.9 | 88.1 | 88.4 |
OpenCompass · HLE | 25.4 | 28.6 | 27.5 |
OpenCompass · IFEval | 90.2 | 93.9 | 91.5 |
OpenCompass · LiveCodeBenchV6 | 83.8 | 80.6 | 83.0 |
OpenCompass · MMLU-Pro | 84.0 | 86.2 | 87.6 |
OTIS Mock AIME 2024-2025 OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills. | 83.3 | 92.2 | 88.9 |
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?" | — | 58.9 | 19.9 |
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. | — | 39.5 | 48.2 |
Artificial Analysis · Quality Index | — | 46.8 | 21.4 |
Chatbot Arena Elo · Coding | 1434.5 | — | 1399.4 |
Chatbot Arena Elo · Overall | 1441.5 | — | 1441.8 |
Cl Bench | 15.9 | 19.3 | — |
Cl Bench Life | 10.9 | 13.2 | — |
FrontierMath-2025-02-28-Private FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning. | 4.3 | 49.0 | — |
FrontierMath-Tier-4-2025-07-01-Private FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning. | 0.0 | 7.0 | — |
PostTrainBench | 7.5 | 10.3 | — |
SimpleBench SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking. | 37.2 | 36.2 | — |
SimpleQA Verified SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information. | 32.2 | 34.3 | — |
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. | 33.4 | 43.2 | — |
Pricing · per 1M tokens · projected $/mo at 10M tokens
| Model | Input | Output | Context | Projected $/mo |
|---|---|---|---|---|
| $0.60 | $2.20 | 205K tokens (~102 books) | $10.00 | |
| $0.45 | $2.25 | 262K tokens (~131 books) | $9.00 | |
| $0.55 | $3.50 | 262K tokens (~131 books) | $12.88 |