Kimi K2.5vsKimi K2 Thinking
Across 11 shared benchmarks, Kimi K2.5 leads overall: Kimi K2.5 wins 7, Kimi K2 Thinking wins 4, with 0 ties and an average score difference of +1.65.
Kimi K2.5
Moonshot AI · 2026-01-27 · Multimodal model
Kimi K2 Thinking
Moonshot AI · 2025-11-06 · Reasoning model
Kimi K2.57 wins(64%)(36%)4 winsKimi K2 Thinking
Benchmark scores
Grouped by capability, sorted by largest gap within each. 11 shared benchmarks.
General Knowledge
Even 4/4| Benchmark | Kimi K2.5 | Kimi K2 Thinking | Diff |
|---|---|---|---|
| LiveBench | 69.0742 / 115Thinking (No Tools) | 61.5963 / 115Thinking (No Tools) | +7.48 |
| MMLU Pro | 78.5069 / 132Thinking (No Tools) | 84.6033 / 132 | -6.10 |
| GPQA Diamond | 87.6037 / 187Thinking (No Tools) | 84.5057 / 187 | +3.10 |
| HLE | 50.2027 / 172Thinking (With Tools) | 5123 / 172 | -0.80 |
Coding and Software Engineer
Kimi K2.5 2/2| Benchmark | Kimi K2.5 | Kimi K2 Thinking | Diff |
|---|---|---|---|
| SWE-bench Verified | 76.8030 / 112Thinking (With Tools) | 71.3055 / 112 | +5.50 |
| LiveCodeBench | 8516 / 123Thinking (No Tools) | 83.1022 / 123 | +1.90 |
Math and Reasoning
Even 2/2| Benchmark | Kimi K2.5 | Kimi K2 Thinking | Diff |
|---|---|---|---|
| FrontierMath - Tier 4 | 4.2040 / 80Normal (No Tools) | 072 / 80Thinking (No Tools) | +4.20 |
| AIME2025 | 96.1021 / 107Thinking (No Tools) | 1001 / 107 | -3.90 |
AI Agent - Information Search
Kimi K2.5 1/1| Benchmark | Kimi K2.5 | Kimi K2 Thinking | Diff |
|---|---|---|---|
| BrowseComp | 60.6036 / 53Thinking (With Tools + Internet) | 60.2037 / 53 | +0.40 |
Claw-style Agent Evaluation
Kimi K2 Thinking 1/1| Benchmark | Kimi K2.5 | Kimi K2 Thinking | Diff |
|---|---|---|---|
| Claw Bench | 81.7018 / 29Thinking (With Tools) | 82.5017 / 29Thinking (With Tools) | -0.80 |
Commonsense Reasoning
Kimi K2.5 1/1| Benchmark | Kimi K2.5 | Kimi K2 Thinking | Diff |
|---|---|---|---|
| Simple Bench | 46.8030 / 63Thinking (No Tools) | 39.6041 / 63Normal (No Tools) | +7.20 |
Specs
| Field | Kimi K2.5 | Kimi K2 Thinking |
|---|---|---|
| Publisher | Moonshot AI | Moonshot AI |
| Release date | 2026-01-27 | 2025-11-06 |
| Model type | Multimodal model | Reasoning model |
| Architecture | MoE | MoE |
| Parameters | 1T | 1T |
| Context length | 256K | 256K |
| Max output | 16K | Not available |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | Kimi K2.5 | Kimi K2 Thinking |
|---|---|---|
| Text input | $0.6 / 1M tokens | $0.6 / 1M tokens |
| Text output | $3 / 1M tokens | $2.5 / 1M tokens |
| Cache read | $0.1 / 1M tokens | $0.3 / 1M tokens |
| Cache write | Not public | $0.6 / 1M tokens |
Summary
- Kimi K2.5leads in:Coding and Software Engineer (2/2), AI Agent - Information Search (1/1), Commonsense Reasoning (1/1)
- Kimi K2 Thinkingleads in:Claw-style Agent Evaluation (1/1)
- Tied in:General Knowledge, Math and Reasoning
On average across the 11 shared benchmarks, Kimi K2.5 scores 1.65 higher.
Largest single-benchmark gap: LiveBench — Kimi K2.5 69.07 vs Kimi K2 Thinking 61.59 (+7.48).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.