Kimi K2.7 Code vs Kimi K3
| Spec | Kimi K2.7 Code | Kimi K3 |
|---|---|---|
| Respan Index | 36.1 (#78) | 52.8 (#35) |
| Blended price per 1M | $1.36 | $2.52 |
| Context window | 262K | 1.0M |
| Input / output per 1M | $0.65 / $3.5 | $0.192 / $9.5 |
| Max output | - | 1.0M |
| Input | text, vision | text, vision |
| Released | 2026-06-12 | 2026-07-16 |
Respan Index by category
Each category is scored 0 to 100, and how far ahead the winner is. Leads under a point are close to even.
| Category | Kimi K2.7 Code | Kimi K3 |
|---|---|---|
| Coding | 12.5 Kimi K2.7 Code | 38.8 Kimi K3 |
| Agents | 49.4 Kimi K2.7 Code | 61.8 Kimi K3 |
| Reasoning | 36.4 Kimi K2.7 Code | 54.8 Kimi K3 |
| Knowledge & language | 36.2 Kimi K2.7 Code | 52.9 Kimi K3 |
| Human preference | 52.9 Kimi K2.7 Code | 57.7 Kimi K3 |
Benchmarks both report
Kimi K3 leads on 24 of the 24 benchmarks both report. Higher is better. Each score links to where it was published.
| Benchmark | Kimi K2.7 Code | Kimi K3 |
|---|---|---|
| LMArena WebDev (LMArena) | 1475 | 1658 |
| Chess Puzzles (Epoch AI) | 14% | 37% |
| APEX-Agents (Mercor) | 37.6% | 47.8% |
| DeepSWE v1.1 (Datacurve) | 30.5% | 68.5% |
| FrontierMath Tier 4 v2 (Epoch AI) | 12.2% | 39% |
| FrontierMath Tiers 1-3 v2 (Epoch AI) | 54% | 72.2% |
| GPQA Diamond | 87.9% | 93.5% |
| LiveBench Agentic Coding (LiveBench) | 45.7% | 62.2% |
| LiveBench Coding (LiveBench) | 74% | 81.4% |
| LiveBench Data Analysis (LiveBench) | 62.7% | 78.7% |
| LiveBench Instruction Following (LiveBench) | 56.3% | 71.4% |
| LiveBench Language (LiveBench) | 77.9% | 85.5% |
| LiveBench Mathematics (LiveBench) | 79.6% | 84.4% |
| LiveBench Reasoning (LiveBench) | 82.8% | 90.7% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 95.6% | 97.2% |
| SimpleQA Verified | 36.5% | 50.6% |
| Vending-Bench 2 (Andon Labs) | 5082.94 | 5165.04 |
| WeirdML (Håvard Tveit Ihle) | 54.1% | 82.6% |
| SimpleBench (SimpleBench) | 57.9% | 60.7% |
| Toolathlon-Verified (HKUST) | 58% | 76.5% |
| MCP-Atlas | 76% | 84.2% |
| MCPMark-Verified | 81.1% | 94.5% |
| MLS-Bench-Lite | 35.1% | 48.3% |
| ProgramBench | 53.6% | 77.8% |
FAQ
Is Kimi K2.7 Code better than Kimi K3?
Kimi K3 scores 52.8 on the Respan Index (#35), ahead of Kimi K2.7 Code at 36.1 (#78). Kimi K3 leads on 24 of the 24 benchmarks both report.
Which is cheaper, Kimi K2.7 Code or Kimi K3?
Kimi K2.7 Code lists at $0.65 input and $3.5 output per 1M tokens; Kimi K3 at $0.192 and $9.5. Kimi K2.7 Code is 1.8x cheaper per token at list price.
Which has the larger context window?
Kimi K2.7 Code accepts 262K tokens and Kimi K3 1.0M tokens.