Kimi K2 Instruct
| Released | 2025-07-11 |
|---|---|
| Price per 1M | $0.50 in · $2.00 out |
| Overall score | 46.8 · rank 140 |
| Context window | 131,072 tokens |
| Provider | Moonshot AI |
| Weights | Open weights (Modified MIT) |
Scores
| Dimension | Score | Rank | Benchmarks |
|---|---|---|---|
| Overall: Balanced | 46.8 | 140 | 10 |
| Overall: Coding | 46.3 | 145 | 7 |
| Overall: Reasoning | 43.6 | 169 | 3 |
| Overall: Agentic work | 46.1 | 143 | 7 |
| Overall: Frontend | 46.6 | 120 | 3 |
| Coding | 46.6 | 143 | 3 |
| Agents | 46.4 | 129 | 2 |
| Reasoning | 44.1 | 161 | 2 |
| Knowledge | 42.5 | 132 | 1 |
| General | 56.7 | 50 | 2 |
| Value vs price | −0.5 | 123 |
Evidence
Coding 46.6 ±2.9
| Benchmark | Result | Source | Date |
|---|---|---|---|
| GSO | 4.9% | Epoch AI | 2026-11-11 |
| Aider Polyglot | 59.1% | Epoch AI | 2026-11-11 |
| LiveCodeBench (Vals run) | 70.4% | Vals AI | 2026-09-01 |
Agents 46.4 ±2.1
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Terminal-Bench 2.0 | 27.8% | Epoch AI | 2025-11-01 |
| BFCL v4 | 59.1% | BFCL | 2026-04-12 |
Reasoning 44.1 ±3.8
| Benchmark | Result | Source | Date |
|---|---|---|---|
| GPQA Diamond (Vals run) | 71.5% | Vals AI | 2026-09-01 |
| SimpleBench | 26.3% | Epoch AI | 2026-11-11 |
Knowledge 42.5 ±4.7
| Benchmark | Result | Source | Date |
|---|---|---|---|
| MMLU-Pro (Vals run) | 79.4% | Vals AI | 2026-09-01 |
General 56.7 ±3.0
| Benchmark | Result | Source | Date |
|---|---|---|---|
| EQ-Bench Creative Writing v3 | 1,666 | EQ-Bench | 2026-11-11 |
| EQ-Bench Longform Creative Writing | 68.1% | EQ-Bench | 2026-11-11 |
Results not used in the score
| Benchmark | Result | Source | Why not used |
|---|---|---|---|
| Terminal-Bench 2.0 | 26.7% | Epoch AI | A stronger configuration result was used |
Names matched to this model
kimi-k2-0711-preview, kimi-k2-instruct, moonshotai-kimi-k2-instruct