2026 DEEPSEEK V4
ПОЛНЫЙ_
РЕЛИЗ_GA.

DeepSeek V4 GA: архитектура CSA+HCA, пик/внепик тарификация, бенчмарки

Тезис: Три месяца preview — и 20 июля 2026 DeepSeek V4 выходит в GA. Agent orchestration, math reasoning, code gen — затюнены под production. Новинка: пик/внепик тарификация (peak-valley pricing). Ниже — полный разбор: timeline, V4-Pro/Flash specs, CSA+HCA+mHC stack, три inference mode, benchmark tables, pricing grid, API migration (24 июля deadline), vs GPT-5.6 / Claude Fable 5, FAQ JSON-LD, deep insight, Mac/Metal angle.

30 секунд · TL;DR

Релиз2026-07-20 GA · deepseek-chat EOL 24.07.
ScaleV4-Pro 1,6T (49B active) · V4-Flash 284B (13B active) · 1M context each
BenchmarksSWE-bench Verified 80,6% (open-weight record) · LiveCodeBench 93,5% · MIT
PricingPeak-valley · V4-Pro output $0,87/M (off-peak) / $1,74/M (peak)
Cost ratio1/116 vs Claude Fable 5 ($0,03 vs $3,48/task)

1. Pain points: зачем смотреть на полный релиз

  1. Preview был силён, production требовал SLA. С 24 апреля 2026 два вопроса: будет ли ещё performance? Изменится ли pricing? GA отвечает — три месяца production tuning + прозрачная peak-valley схема.
  2. Legacy API умирает через четыре дня. deepseek-chat и deepseek-reasoner отключаются 24.07.2026, 15:59 UTC. Cursor, OpenClaw, custom agents с старыми model strings — мигрируйте сейчас.
  3. Peak-valley — инструмент load shaping, не просто price hike. Пекинские будни 09–12 и 14–18 — rates ×2. Но даже в peak V4-Pro output ($1,74/M) — 8,6× дешевле Opus 4.8 ($15/M). Ключ: off-peak scheduling + Flash routing.

2. Что это: DeepSeek V4 GA в одном абзаце

DeepSeek V4 GA — general availability от 20.07.2026: V4-Pro (1,6T MoE) и V4-Flash (284B MoE), MIT license, 1M token context window, 384K max output.

Vs preview: agent capabilities, math reasoning, code generation — production-tuned; peak-valley API pricing. Архитектура CSA + HCA + mHC — с апрельского preview; GA закрывает stability, tuning и commercial loop.

One-liner: лучший SWE-bench Verified среди open-weight (80,6%), near-frontier performance за 1/10–1/100 стоимости closed flagship, private deploy + реальный 1M context.

3. Timeline: preview → GA

ДатаСобытие
2026-04-24V4 preview + open source (MIT), V4-Pro (1,6T) и V4-Flash (284B)
2026-05Production-tuned builds, API GA
2026-06V4-Pro permanent −75% (output $0,87/M)
2026-06-29Email всем API users: GA mid-July, первое раскрытие peak-valley
2026-07-19GA greyscale access, медиа «полный релиз завтра»
2026-07-20GA live (дата этой статьи)
2026-07-24deepseek-chat / deepseek-reasoner permanent EOL

Контекст: DeepSeek custom chip & inference, обзор Kimi K3, AI скидки июнь 2026.

4. Model family: V4-Pro vs V4-Flash

SpecV4-ProV4-Flash
Total params1,6T284B
Active params/token49B13B
Transformer layers6143
Context window1 000 000 tokens1 000 000 tokens
Max output384K tokens384K tokens
PrecisionFP4 (experts) + FP8 (rest)FP4 + FP8 mixed
Pretrain data33T+ tokens32T+ tokens
LicenseMITMIT

5. Архитектура: CSA + HCA + mHC — engineering deep dive

Классический Transformer: KV Cache растёт O(n) с context length — 1M tokens на inference silicon было экономически нереально. V4 выбрасывает MLA из V2/V3 и ставит три архитектурных хода:

5.1 CSA — Compressed Sparse Attention

  • KV sequence сжимается через Softmax-gated pooling
  • FP4 «Lightning Indexer» для top-k sparse selection (V4-Pro: top-1024, V4-Flash: top-512)
  • Sliding window последних 128 tokens
  • На 1M context: 27% inference FLOPs vs DeepSeek V3.2 — это прямой выигрыш в tokens/sec на inference ASIC и Metal GPU alike

5.2 HCA — Heavily Compressed Attention

  • Token compression 128×, global dense attention
  • Long-range dependencies — complement к CSA
  • V4-Flash: первые 2 слоя HCA, далее CSA/HCA alternate; V4-Pro — аналогично

Net effect: KV Cache memory = 10% от V3.2 на 1M (V4-Flash: 7%). Меньше HBM pressure → выше batch throughput на фиксированном silicon budget.

5.3 mHC + Muon optimizer

mHC (Manifold-Constrained Hyper-Connections) — upgrade residual path: x = x + f(x) заменяется 4-channel residual stream с mixing matrices на Birkhoff polytope (doubly-stochastic constraint). 61-layer deep network без gradient pathology.

Training: Muon optimizer вместо AdamW — Newton-Schulz orthogonalization градиентов, faster convergence, stable training at 1,6T scale.

5.4 Три inference mode

ModeХарактеристикаUse case
Non-thinkNo CoT, min latencySimple Q&A, routing
Think HighExplicit reasoning (<redacted_thinking>)Mid complexity, code debug
Think MaxMax reasoning depth, 384K+ contextHard math, long agent chains

Sampling defaults: temperature=1.0, top_p=1.0 (official, all modes).

6. Benchmarks: open-weight leaderboard

BenchmarkDeepSeek V4-ProClaude Fable 5GPT-5.6 UltraClaude Opus 4.8
SWE-bench Verified80,6% ★ open-weight #196,0%not separately published~69%
SWE-bench Pro55,4%80,3%78,1%69,2%
LiveCodeBench (Pass@1)93,5%88,1%87,4%83,2%
Codeforces Elo3 206
Terminal-Bench 2.183,9%88,0%85,1%82,7%

Readout:

  • SWE-bench Verified: V4-Pro 80,6% — open-weight top, tie с Gemini 3.1 Pro
  • LiveCodeBench: V4-Pro 93,5% — ahead всех closed competitors
  • SWE-bench Pro: Fable 5 80,3% — multi-file repo bugfixes
  • Terminal-Bench: GPT-5.6 Ultra 85,1% — terminal/toolchain-heavy agents

Disclaimer: benchmarks — DeepSeek self-reported + third-party aggregation; independent reproduction in progress. Harness mismatch между моделями — учитывайте при сравнении.

7. Cost-per-task: economics at scale

Artificial Analysis, Strategy & Ops Index:

ModelCost/taskScoreRatio
Claude Fable 5$3,4850 ptsbaseline
DeepSeek V4-Pro$0,0338 pts1/116 Fable 5 cost
DeepSeek V4-Flashall 6 index tasks < $0,04light tasks optimal

Fable 5 = 116× дороже за ~12 pts (31%) прироста. Для большинства production workloads V4-Pro — rational default.

8. Head-to-head: V4 vs GPT-5.6 Sol vs Claude Fable 5

DimensionDeepSeek V4-ProGPT-5.6 SolClaude Fable 5
Open source✅ MIT❌ closed❌ closed
Self-host
Context1M tokensundisclosed1M tokens
Codevery strong (near Fable 5)very strongstrongest tier
API output (off-peak)$0,87/M~$15/M$50/M
API output (peak)$1,74/M
Agent capabilitystrong, multi-agentstrongstrongest tier
Data privacy✅ private deploy
ScenarioPickWhy
Budget / high-frequency / self-hostV4-Pro / Flash$0,87/M, MIT
Max code quality, cost irrelevantClaude Fable 5SWE-bench Pro 80,3%
Algorithms + math reasoningGPT-5.6 Sol / UltraTerminal-Bench lead
Massive log/doc processingV4-Flashcache hit input $0,0028/M

9. Peak-valley pricing: full rate card

GA впервые вводит time-of-use differentiation — как electricity TOU rates. Peak (Пекин): будни 09:00–12:00 и 14:00–18:00, все rates ×2.

ModelLine itemOff-peakPeak
V4-ProInput (cache hit)¥0,025 / $0,0035 / 1M×2
Input (cache miss)¥3,00 / $0,435 / 1M¥6,00 / $0,87
Output¥6,00 / $0,87 / 1M¥12,00 / $1,74
V4-FlashInput (cache hit)¥0,02 / $0,0028 / 1M×2
Input (cache miss)¥1,00 / $0,14 / 1M¥2,00 / $0,28
Output¥2,00 / $0,28 / 1M¥4,00 / $0,56

9.1 Четыре cost hacks

  1. Non-realtime → off-peak: doc batches, annotation, code review после 18:00 или до 09:00
  2. Prompt cache: repeated system prompts — V4-Flash cache hit от $0,0028/M
  3. Flash routing: intent на Flash, hard reasoning на Pro — 60–80% cost cut
  4. Billing alert: DeepSeek promises 24h email notice перед rate changes

Даже в peak: V4-Pro ($1,74/M) — 8,6× cheaper Opus 4.8 ($15/M) и GPT-5.6 Sol (~$15/M).

10. API migration (deadline 24 июля) ⚠️

Hard stop: deepseek-chat и deepseek-reasoner EOL 24.07.2026, 15:59 UTC (23:59 Пекин).

LegacyNewNote
deepseek-chatdeepseek-v4-flash (non-think)Fast, light tasks
deepseek-reasonerdeepseek-v4-flash (think)or deepseek-v4-pro for stronger reasoning

10.1 OpenAI SDK (Python)

from openai import OpenAI client = OpenAI( api_key="your-deepseek-api-key", base_url="https://api.deepseek.com/v1" ) # ❌ Legacy (dead after 24.07.) # client.chat.completions.create(model="deepseek-chat", messages=[...]) # ✅ New — non-think response = client.chat.completions.create( model="deepseek-v4-flash", messages=[{"role": "user", "content": "Hello, DeepSeek V4!"}] ) # ✅ New — think (replaces deepseek-reasoner) response = client.chat.completions.create( model="deepseek-v4-pro", messages=[{"role": "user", "content": "Solve this step by step..."}], extra_body={ "thinking": {"type": "enabled", "budget_tokens": 8000} } )

10.2 Anthropic SDK

V4 supports OpenAI ChatCompletions + Anthropic Messages — update model only:

import anthropic client = anthropic.Anthropic( api_key="your-deepseek-api-key", base_url="https://api.deepseek.com" ) message = client.messages.create( model="deepseek-v4-pro", max_tokens=4096, messages=[{"role": "user", "content": "Что изменилось в V4 GA release?"}] )

11. Integration: пять путей в production

  1. Web/App: chat.deepseek.com — V4 GA default.
  2. Official API: platform.deepseek.com — OpenAI-compatible, deepseek-v4-pro / deepseek-v4-flash.
  3. OpenRouter: model ID deepseek/deepseek-v4-pro — см. OpenRouter Mac routing guide.
  4. Cursor / IDE: V4-Pro daily driver, fallback Claude Fable 5 для hard repo bugs.
  5. Codebase grep: search deepseek-chat + deepseek-reasoner, staging test, prod cutover до 24 июля.

12. FAQ

Q: GA vs preview?
A: Agent/math/code improvements, peak-valley pricing, production stability. Architecture unchanged.

Q: V4-Pro vs V4-Flash?
A: Pro — complex reasoning, long agents; Flash — high-frequency routing ($0,28/M off-peak).

Q: Peak hours?
A: Пекин будни 09:00–12:00 и 14:00–18:00, rates ×2.

Q: deepseek-chat EOL?
A: 24.07.2026 15:59 UTC. Migrate to deepseek-v4-flash or deepseek-v4-pro.

Q: Self-host / local on Mac?
A: MIT, Hugging Face weights. V4-Pro needs GPU cluster; Mac for API or quantized V4-Flash via MLX.

Q: vs Claude Fable 5?
A: Fable 5 leads SWE-bench; V4-Pro at 1/116 cost with near-frontier perf + self-host.

13. Deep insight: open-source milestone signal 2026

DeepSeek V4 GA — не «preview graduation». Это structural inflection point в китайском open-source LLM ecosystem mid-2026.

13.1 От price disruptor к disciplined platform

18 месяцев DeepSeek держал рынок на ~$0,28/M output (V3 era), forcing OpenAI/Anthropic price cuts. «Почти бесплатно» unsustainable — compute, R&D, compliance требуют cash flow. Peak-valley = signal от subsidy → sustainable ops. Competitiveness не sacrificed: peak $1,74/M всё ещё 8,6× under Opus 4.8. Mechanism = demand-side scheduling, не cost passthrough. Teams с off-peak batch pipelines могут платить меньше, чем раньше.

13.2 Architecture redefines long-context economics

1M context в 2026 — commodity spec (Kimi K3, Fable 5). V4 value prop: 1M context economically servable. CSA 4× + top-1024 + 128-token window → 27% FLOPs, 10% KV Cache vs V3.2. Kimi K3 $15/M output vs V4-Pro $0,87/M — order-of-magnitude gap на million-token workloads.

mHC 4-channel residual + Muon = stable training 61 layers at 1,6T. Parameter jump 671B → 1,6T без training collapse.

13.3 MIT license — strategic weight

July 2026: license = parameter count в importance. Llama commercial restrictions, «research only» clauses — compliance landmines. MIT на V4:

  • Unlimited on-prem, no royalty, no reporting
  • Finetune, distill, ship closed-source products
  • Data residency для gov/finance/health без US API egress

Kimi K3 claims 2,8T «largest open» (см. K3 обзор); V4 competes on engineering efficiency, cost structure, license freedom. Для Mac dev: MIT + dual SDK (OpenAI/Anthropic) = seamless switch в Cursor, Continue, OpenClaw; MLX/llama.cpp quantization path open. V4-Flash 13B active на Apple Silicon unified memory — community quant experiments running; track Apple Silicon V4 AMX-2 benchmarks для local tokens/sec и Metal throughput numbers.

Bottom line: V4 GA не бьёт Fable 5 / GPT-5.6 на every leaderboard — yet. Но delivers undisputed strongest open-weight performance at 1/10–1/100 closed flagship cost. Для teams без data egress, budget-constrained devs, 1M-context agent engineers — V4-Pro = no-compromise pick.

14. Closing: API anywhere — agent validation on Mac/Metal

platform.deepseek.com API key, OpenRouter route swap — Windows/Linux достаточно. Но если вам нужен Cursor + V4-Pro long-context coding, Think Max stress на multi-repo SWE-bench regression, 1M doc batch jobs, или MLX quantized V4-Flash validation перед prod deploy — Apple Silicon unified memory + Metal compute stack = lowest-friction path. Zero PCIe bottleneck на KV-heavy workloads vs discrete GPU + system RAM split.

Pragmatic split: daily driver на API, V4-Pro Think Max load tests, SWE-bench regression suites, 1M context doc batches — на MACGPU remote Mac mini M4 node. On-demand, SSH-secured, peak-valley cron scheduling на remote silicon. Main machine остаётся stable для everyday dev; Metal throughput на remote node съедает agent memory pressure.