{"schema":"postcutoff/event@1","as_of":"2026-11-13T13:49:00+01:00","url":"https://postcutoff.com/e/2026-04-24-deepseek-v4-preview/","md":"https://postcutoff.com/e/2026-04-24-deepseek-v4-preview/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"2026-04-24-deepseek-v4-preview","date":"2026-04-24","date_precision":"day","short_title":"DeepSeek V4 preview: a 1.6T-parameter open MoE running on Huawei Ascend","deck":null,"takeaway":"DeepSeek released a preview of V4 on 2026-04-24: V4-Pro (1.6T total / 49B active) and V4-Flash (284B / 13B active), both MIT-licensed MoE models with a 1M-token context, validated on Huawei Ascend NPUs as well as Nvidia GPUs, priced far below Western frontier APIs.","category":"model-release","category_label":"Model releases","importance":5,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":[{"n":1,"title":"DeepSeek API changelog","url":"https://api-docs.deepseek.com/updates/","type":"docs","group":"primary","domain":"api-docs.deepseek.com"},{"n":2,"title":"The Register: DeepSeek's new models offer big inference cost savings","url":"https://www.theregister.com/2026/04/24/deepseek_v4/","type":"press","group":"press","domain":"theregister.com"},{"n":3,"title":"Tom's Hardware: DeepSeek launches 1.6T V4 on Huawei chips","url":"https://www.tomshardware.com/tech-industry/artificial-intelligence/deepseek-launches-1-6-trillion-parameter-v4-on-huawei-chips-as-us-escalates-ai-theft-accusations","type":"press","group":"press","domain":"tomshardware.com"},{"n":4,"title":"Huawei Central: DeepSeek launches V4 on Huawei chips","url":"https://www.huaweicentral.com/deepseek-launches-new-v4-ai-models-running-on-huawei-chips/","type":"press","group":"press","domain":"huaweicentral.com"}],"official":1,"filed":"2026-09-29","updated":"2026-09-29","orgs":["DeepSeek"],"title":"DeepSeek V4 preview: 1.6T-parameter open MoE running on Huawei Ascend","summary":"DeepSeek released a preview of V4 on 2026-04-24: V4-Pro (1.6T total / 49B active) and V4-Flash (284B / 13B active), both MIT-licensed MoE models with a 1M-token context, validated on Huawei Ascend NPUs as well as Nvidia GPUs, priced far below Western frontier APIs.","key_facts":["V4-Pro: 1.6T total parameters, 49B active; V4-Flash: 284B total, 13B active (The Register)","Training data: 33T tokens; context window 1M tokens","KV cache 9.5x-13.7x smaller than DeepSeek V3.2; mixed FP8/FP4 precision with quantization-aware training of MoE experts","New hybrid attention (Compressed Sparse Attention + Heavily Compressed Attention) and Muon optimizer","API price: Flash $0.14/M input, $0.28/M output; Pro $1.74/M input, $3.48/M output","Day-zero support on Huawei Ascend SuperNode line incl. Ascend 950; weights on Hugging Face under MIT license"],"key_numbers":[],"tags":["llm","open-weights","china","moe","huawei-ascend","long-context"],"science":null,"body_md":"## What happened\n\nOn Friday 2026-04-24 DeepSeek published a **preview** of its fourth-generation model family. Two MoE models shipped:\n**V4-Pro** (1.6 trillion parameters, 49B active) and **V4-Flash** (284B, 13B active), both with a 1M-token context window and trained on ~33T tokens.\nArchitecturally, DeepSeek introduced a hybrid compressed attention scheme and adopted the Muon optimizer, and cut KV-cache memory 9.5-13.7x versus V3.2,\nusing FP8/FP4 mixed precision with quantization-aware training.\n\nThe launch was notable for hardware: DeepSeek validated the models on **Huawei Ascend** NPUs (Huawei announced day-zero support across its SuperNode line,\nincluding Ascend 950) as well as Nvidia GPUs. Coverage (Tom's Hardware) linked the release to escalating US government accusations of IP theft / distillation by Chinese labs.\nLater milestones: V4-Flash re-post-trained update (2026-07-31), V4-Pro GA with low/high/max thinking effort (2026-08-13), and V4.1-Flash (2026-09-10).\n\n## Why it matters\n\nV4 was the largest open-weights model at release and the first frontier-class release optimized for a Chinese AI accelerator, a signal that China's\nmodel stack can decouple from Nvidia. Its aggressive pricing (Pro output $3.48/M) kept pressure on Western API prices.","disputed":[],"related":[{"id":"2026-09-18-huawei-ascend-950-cluster-cloud","url":"https://postcutoff.com/e/2026-09-18-huawei-ascend-950-cluster-cloud/","date":"2026-09-18","date_precision":"day","short_title":"Huawei sets Ascend 950 cluster cloud launch and Ascend 960 roadmap","deck":null,"takeaway":"Ascend 950 is China's main answer to US export controls; selling it as a global cloud service extends Huawei's AI compute beyond China.","category":"hardware-compute","category_label":"Chips & compute","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":5,"official":0,"filed":"2026-09-29","updated":"2026-10-09","orgs":["Huawei"]},{"id":"2026-09-10-deepseek-v4-1-flash","url":"https://postcutoff.com/e/2026-09-10-deepseek-v4-1-flash/","date":"2026-09-10","date_precision":"day","short_title":"DeepSeek V4.1-Flash brings a new architecture family, native vision and a cheaper API","deck":null,"takeaway":"The \"new architecture family\" framing implies larger V4.1 models are coming.","category":"model-release","category_label":"Model releases","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":10,"official":5,"filed":"2026-09-29","updated":"2026-10-18","orgs":["DeepSeek"]}],"people":[],"posts":[],"videos":[],"models":[],"changes":[{"date":"2026-09-29","type":"filed","text":"Created"}],"provenance":{"agents":[{"model":"Claude Opus 5.5","maker":"Anthropic","tool":"Claude Code"}],"filed":"2026-09-29","run":null,"sources_read":null,"updated":"2026-09-29","human_review":null,"version":{"date":"2026-10-07"}},"gaps":[{"model_id":"gpt-6-astra","name":"GPT-6 Astra","cutoff":"2026-04","days_after":null,"in_training_data":true},{"model_id":"claude-opus-5-5","name":"Claude Opus 5.5","cutoff":"2026-06","days_after":null,"in_training_data":true},{"model_id":"gemini-4-hydrogen","name":"Gemini 4 Hydrogen","cutoff":"2026-03","days_after":24,"in_training_data":false},{"model_id":"grok-4-7","name":"Grok 4.7","cutoff":"2026-05","days_after":null,"in_training_data":true}],"short_url":"https://postcutoff.com/s/deepseek-v4-preview"}