Ночью 16 июля 2026 Moonshot AI включила Kimi K3 без keynote — баннер в API-доке, tech blog и рабочий endpoint. Этот обзор для ML-инженеров, SRE и agent-разработчиков, которым нужны не маркетинговые тезисы, а цифры: 2,8 трлн параметров, механизм Kimi Delta Attention, сравнение с Claude Fable 5 / GPT-5.6 Sol, цены API, пошаговая интеграция и дата open weights — 27 июля 2026.
01 Kimi K3: параметры, MoE routing и production bottlenecks
Kimi K3 — крупнейшая open-source LLM на сегодня: 2,8T параметров (+75 % к DeepSeek V4 Pro 1,6T). Sparse MoE: 896 experts, на forward pass активируются 16 (sparsity 1,8 %). Контекст: 1 048 576 tokens. Native vision (image/video). Фокус — long-horizon coding и knowledge work.
| Параметр | Значение | Контекст |
|---|---|---|
| Total params | 2,8T | Record open source |
| MoE config | 896 / 16 active | Expert routing 1,8 % |
| Context window | 1 048 576 | 5× Claude Sonnet 5 |
| Multimodal | Vision native | OmniDocBench 91,1 |
| Model ID | kimi-k3 | OpenAI-compatible API |
| Open weights | 27.07.2026 | Hugging Face — full checkpoint |
Стратегический контекст: Kimi лидировала по размеру open models 9 из 12 месяцев; релиз — накануне WAIC 2026. ARR >$300M (июнь 2026), valuation $31,5B pre-money, API revenue >70 %, international paid users +400 %.
Production pain points, которые specs не снимают:
- Cloud lock-in до 27.07.: pipeline зависит от Moonshot infra — latency, quotas, pricing drift.
- 1M context ≠ zero engineering cost: без persistent dev environment (bare-metal Mac, стабильный toolchain) agent resume и tool integration ломаются.
- Self-host barrier: Moonshot указывает 64+ accelerators для production inference — consumer hardware out of scope.
- Benchmark harness mismatch: Kimi Code vs Claude Code vs Codex — независимые reproductions in progress.
- Metal/Xcode stack: API-модель не заменяет native Apple Silicon для iOS CI/CD и long-running agents на физическом Mac.
02 Архитектура: KDA scheduling, AttnRes и MoE stability
K3 — не naive scaling. Сравните с моделями из OpenRouter June ranking: Moonshot оптимизирует long-context decode и training stability при extreme sparsity.
Kimi Delta Attention (KDA)
Full attention — O(n²); при 1M tokens KV-cache съедает HBM. KDA чередует linear и full attention в ratio 3:1:
- 3 linear layers — cheap local structure; 1 full layer — global information flow;
- KV-cache memory: до −75 %;
- Decode at 1M tokens: до 6,3× быстрее;
- Beat full-attention baseline на short/long context и RL scaling.
Attention Residuals (AttnRes)
- Standard residual path размывает early-layer representations на глубине;
- AttnRes — selective retrieval high-value features через depth skip;
- Training efficiency +25 %, compute overhead <2 %.
Stable LatentMoE
| Technique | Function |
|---|---|
| Quantile Balancing | Expert assignment из router quantiles — без heuristic hyperparams |
| Per-Head Muon | Per-head optimizer для stable large-scale training |
| Sigmoid Tanh Unit (SiTU) | Activation control |
| Gated MLA | Attention selectivity |
| Net effect | ~2,5× scaling efficiency vs Kimi K2 |
03 Benchmark matrix: K3 vs Claude Fable 5 vs GPT-5.6 Sol
Self-report Moonshot, 16.07.2026. Для repo-level bugfix — Claude Code на Mac mini.
| Benchmark | Kimi K3 | Claude Fable 5 | GPT-5.6 Sol | Claude Opus 4.8 |
|---|---|---|---|---|
| DeepSWE | 67,5 | 70,0 | 73,0 | 59,0 |
| Program Bench | 77,8 | 76,8 | 77,6 | 71,9 |
| Terminal Bench 2.1 | 88,3 | 84,6 | 88,8 | 84,6 |
| FrontierSWE | 81,2 | 86,6 | 71,3 | 66,7 |
| SWE Marathon | 42,0 | 35,0 | 39,0 | 40,0 |
| BrowseComp | 91,2 | 88,0 | 90,4 | 84,3 |
| Automation Bench | 30,8 | 29,1 | 29,7 | 27,2 |
| GPQA-Diamond | 93,5 | 92,6 | 94,1 | 91,0 |
| MMMU-Pro | 81,6 | 81,2 | 83,0 | 78,9 |
| OmniDocBench | 91,1 | 89,8 | 85,8 | 87,9 |
Geek read: K3 лидирует SWE Marathon (42,0) — multi-hour coding sessions. Program Bench 77,8 — #1. FrontierSWE — Fable 5 territory (86,6). AA Intelligence Index v4.1: K3 57,1, rank #4.
Self-report, разные harnesses — ждите independent reproduction перед production cutover.
04 API pricing и 6 шагов интеграции
| Model | Input | Output | Cache hit | Context |
|---|---|---|---|---|
| Kimi K3 | $3,00 | $15,00 | $0,30 | 1M |
| Claude Sonnet 5 | $3,00 | $15,00 | — | 200K |
| Claude Opus 4.8 | $5,00 | $25,00 | — | 200K |
| GPT-5.5 | $5,00 | $30,00 | — | 400K |
| DeepSeek V4 Pro | $1,74 | $3,48 | $0,145 | 128K |
| Kimi K2.6 | $0,95 | $4,00 | $0,16 | 256K |
K3 = Sonnet 5 pricing ($3/$15), но 5× context. Cache hits от $0,30/M; Moonshot claims 90 %+ cache hit в coding. China: ¥20/¥100/¥2 per M. Consumer kimi.com — free tier.
6 шагов интеграции:
- Регистрация Moonshot: platform.kimi.ai или kimi.com.
- API key: generate в dashboard, store securely (one-time display).
- Model ID:
kimi-k3; OpenRouter:moonshotai/kimi-k3. - OpenAI-compatible client:
base_url = https://api.moonshot.ai/v1. - Smoke test: short prompt — latency и quota check.
- Cache-aware workflow: reuse system prompts и tool definitions для 90 %+ cache hits.
from openai import OpenAI
client = OpenAI(
api_key="your_moonshot_api_key",
base_url="https://api.moonshot.ai/v1"
)
response = client.chat.completions.create(
model="kimi-k3",
messages=[{"role": "user", "content": "Разбери этот код..."}]
)
print(response.choices[0].message.content)
05 Decision matrix и open weights 27.07.2026
| Scenario | Pick | Why |
|---|---|---|
| Long-horizon coding | Kimi K3 | SWE Marathon #1, max context |
| Repo-level bugfix | Claude Fable 5 | FrontierSWE lead |
| Terminal/tool agents | GPT-5.6 Sol | Terminal Bench 2.1 |
| Long docs + multimodal | Kimi K3 | OmniDocBench 91,1 |
| Cost-sensitive | DeepSeek V4 Pro | Output $3,48/M |
| Self-host post-27.07. | Kimi K3 | Strongest open weights to date |
Moonshot: full weights on Hugging Face 27 июля 2026 — first open model >2T. Training: MXFP4 weights + MXFP8 activations. Expect vLLM/SGLang support и MXFP4/NVFP4 quants. Timeline: WAIC 17–20.07 → 27.07. open weights.
06 Hard data, sources и production conclusion
- Total params: 2,8T
- MoE: 896 experts, 16 active (1,8 % sparsity)
- Context: 1 048 576 tokens
- KDA: KV −75 %, decode +6,3× at 1M
- AttnRes: training +25 %, overhead <2 %
- Scaling vs K2: ~2,5×
- API: $3 / $15 / $0,30 per M tokens
- China: ¥20 / ¥100 / ¥2
- ARR Jun 2026: >$300M; valuation $31,5B
- AA Index v4.1: 57,1 (#4)
- Open weights: 27.07.2026 (Hugging Face)
Re-verify links after upstream updates:
Official:
Moonshot AI — Kimi K3 tech blog
Kimi API Platform — docs and pricing
Third-party:
Artificial Analysis — Intelligence Index v4.1
OpenRouter — moonshotai/kimi-k3
Kimi K3 поднимает планку open source — но pure API = latency, quotas, fragile toolchain; self-host до 27.07. impossible, после — supernode GPU farm; shared cloud VM режет Metal performance и ломает long-running agents. Для native Apple Silicon, 7×24 uptime, stable iOS CI/CD (Kimi Code / Claude Code на persistent physical Mac) — ZUKCLOUD bare-metal Mac mini cloud nodes: dedicated hardware, zero hypervisor tax, flexible billing. Pricing, Order, bare-metal manifesto.
07 FAQ
Q: Kimi K3 бесплатен?
A: kimi.com — free consumer tier (max reasoning). API: $3/$15 per M tokens.
Q: Local deploy?
A: Weights с 27.07.2026 на Hugging Face. Production: 64+ accelerators.
Q: K3 vs DeepSeek V4 Pro?
A: K3 — 2× params, 5× context, stronger coding benches. DeepSeek — output $3,48/M. См. DeepSeek chip ecosystem.
Q: Зачем 1M context?
A: Whole-repo analysis, long legal/scientific docs, agent memory. Flat pricing + high cache rate.
Q: Когда low/high reasoning?
A: Moonshot анонсировала low и high для future updates; сейчас только max.