Kimi K2 Instruct kimi-k2-instruct
moonshotai · kimi flagship agentic
#61 overall #3 coding #9 agentic
- params
- 1000B (A32B)
- arch
- moe
- context
- 128k
- license
- modified-mit open weights
- released
- Jul 2025
- train compute
- 3.0e+24 FLOP
- downloads/30d
- 325.1k
# architecture
- layers
- 61
- d_model
- 7168
- heads
- 64q / 64kv
- head dim
- 112
- experts
- top-8 of 384 + 1 shared
- vocab
- 163840
- family
- kimi_k2
signal path of one layer, generated from the registry's structured fields — dimension lines quote the model's real numbers; an MoE trace forks at the router, a dense trace runs straight through. Geometry fields come from the repo's config.json.
# why ranked
Overall: #61 — score 41.8 ● high — all signals present
| signal | weight | input (0–100) |
|---|---|---|
| aa_intelligence | 0.70 | 45.5 |
| bench_composite | 0.30 | 33.3 |
benchmark panel evidence:
| complete panel | score (0–100) |
|---|---|
| aa-math-v1 | 25.0 |
| aider-polyglot-v1 | 46.8 |
| aime-2025-v1 | 0.0 |
| swe-bench-verified-v1 | 61.2 |
Coding: #3 — score 67.3 ● medium — one signal missing, weights renormalized
| signal | weight | input (0–100) |
|---|---|---|
| aa_coding | 0.30 | — |
| aider_polyglot | 0.30 | 72.7 |
| swe_bench_verified | 0.40 | 63.2 |
missing signals are dropped and the remaining weights renormalized — never imputed.
Agentic: #9 — score 63.2 ● low — a single signal; treat with caution
| signal | weight | input (0–100) |
|---|---|---|
| swe_bench_verified | 0.60 | 63.2 |
| swe_rebench | 0.40 | — |
missing signals are dropped and the remaining weights renormalized — never imputed.
# trend
methodology changed during this history window; score movement across that boundary is not model movement. See methodology v6.
# benchmarks
| benchmark | score | source | date |
|---|---|---|---|
| AA Intelligence Index | 12.7 | Artificial Analysis | — |
| AA Math Index | 57.0 | Artificial Analysis | — |
| Aider Polyglot | 59.1 | Aider | Jul 2025 |
| Acebench | 0.8 / 1 | LLM Stats | — |
| Aider-polyglot | 0.6 / 1 | LLM Stats | — |
| AIME 2024 | 0.7 / 1 | LLM Stats | — |
| AIME 2025 | 0.5 / 1 | LLM Stats | — |
| Autologi | 0.9 / 1 | LLM Stats | — |
| Cnmo-2024 | 0.7 / 1 | LLM Stats | — |
| Gsm8k | 1.0 / 1 | LLM Stats | — |
| Math-500 | 1.0 / 1 | LLM Stats | — |
| SWE-bench Verified | 65.4 | SWE-bench | Jul 2025 |
# where to run
| provider | quant | ctx | $/M in | $/M out | $/M cache | price src | tps | uptime | ✓ |
|---|---|---|---|---|---|---|---|---|---|
| Featherless | — | — | — | — | via hfrouter | — | — | ||
| DeepInfra | 128k | $0.50 | $2.00 | — | via litellm | — | — | ||
| Novita | fp8 | 128k | $0.57 | $2.30 | — | novita | — | 100.0% | ✓ |
| Together | — | $1.00 | $3.00 | — | via litellm | — | — | ||
| Hyperbolic | 128k | $2.00 | $2.00 | — | via litellm | — | — |
sorted by blended price ((3·input + output) / 4 per 1M) · ✓ = the provider's own catalog confirms the offer · "via …" prices are what the aggregator routing the offer charges, not the provider's own list price
# variants
source aliases
- aa
kimi-k2- aider
Kimi K2- arena
Kimi-K2-Instruct- epoch
Kimi K2- openrouter
moonshotai/kimi-k2- swebench
moonshot/kimi-k2-0711-preview