Kimi K3vsKimi K2.5
Across 4 shared benchmarks, Kimi K3 leads overall: Kimi K3 wins 4, Kimi K2.5 wins 0, with 0 ties and an average score difference of +15.52.
Kimi K3
Moonshot AI · 2026-07-16 · Reasoning model
Kimi K2.5
Moonshot AI · 2026-01-27 · Multimodal model
Kimi K34 wins(100%)(0%)0 winsKimi K2.5
Benchmark scores
Grouped by capability, sorted by largest gap within each. 4 shared benchmarks.
General Knowledge
Kimi K3 2/2| Benchmark | Kimi K3 | Kimi K2.5 | Diff |
|---|---|---|---|
| GPQA Diamond | 93.508 / 187最高（无工具） | 87.6037 / 187Thinking (No Tools) | +5.90 |
| HLE | 5610 / 170Max (With Tools) | 50.2025 / 170Thinking (With Tools) | +5.80 |
AI Agent - Information Search
Kimi K3 1/1| Benchmark | Kimi K3 | Kimi K2.5 | Diff |
|---|---|---|---|
| BrowseComp | 91.201 / 52Max (With Tools + Internet) | 60.6035 / 52Thinking (With Tools + Internet) | +30.60 |
AI Agent - Tool Usage
Kimi K3 1/1| Benchmark | Kimi K3 | Kimi K2.5 | Diff |
|---|---|---|---|
| MCP-Atlas | 84.202 / 27Max (With Tools) | 64.4019 / 27Normal (With Tools) | +19.80 |
Specs
| Field | Kimi K3 | Kimi K2.5 |
|---|---|---|
| Publisher | Moonshot AI | Moonshot AI |
| Release date | 2026-07-16 | 2026-01-27 |
| Model type | Reasoning model | Multimodal model |
| Architecture | MoE | MoE |
| Parameters | 2.8T | 1T |
| Context length | 1M | 256K |
| Max output | 1M | 16K |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | Kimi K3 | Kimi K2.5 |
|---|---|---|
| Text input | ¥20 / 1M tokens | $0.6 / 1M tokens |
| Text output | ¥100 / 1M tokens | $3 / 1M tokens |
| Cache read | ¥2 / 1M tokens | $0.1 / 1M tokens |
Summary
- Kimi K3leads in:General Knowledge (2/2), AI Agent - Information Search (1/1), AI Agent - Tool Usage (1/1)
On average across the 4 shared benchmarks, Kimi K3 scores 15.52 higher.
Largest single-benchmark gap: BrowseComp — Kimi K3 91.20 vs Kimi K2.5 60.60 (+30.60).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.