DeepSeek Timeline: Every Model Release and Update

HKUST trained small models inside DeepSeek Harness and nearly tripled a Text-to-SQL agent's score.

DeepSeek's official status page reported a partial outage on the afternoon of November 3.

DeepSeek wants agents that stay online across sessions and run tasks in parallel; an internal instance already spots bugs in group chats and opens PRs.

DeepSeek Harness test code lists three placeholder identifiers, including deepseek-v5, but DeepSeek has not announced V5.

DeepSeek has released the inference framework behind V4.1-Flash, plus a companion repo that replicates how the official API processes tokens.

US federal civilian agencies must remove existing DeepSeek products and stop buying new ones, and the ban also covers third-party services built with DeepSeek.

Unsloth released a 1-bit GGUF of DeepSeek's 552B model that runs locally on machines with at least 128GB of memory.

The October 17 V4.1-Flash update ranks No. 1 among open models and No. 11 overall on Arena's Agent Arena, at a median $0.20 per task.

The open-source dsx engine ships two dedicated runners for the 552B MoE + 196B Engram model: one for a desktop with two RTX 3090s, one for a single AMD MI325X.

The open-source engine Xolotl runs the full 552B DeepSeek-V4.1-Flash on one gaming PC. Its author measured 26.8 tok/s decode.

PR #1225 for antirez/ds4 (dwarfstar) adds support for the new 1017 checkpoint and fixes a check that let mismatched vision towers load.

DeepSeek shipped an updated Flash model with open weights and, for the first time, opened its in-house search engine to outside developers.

DeepSeek V4.1 Flash calls inside OpenCode no longer count against plan credits; the free version tops out at 200K tokens of context.

On the morning of October 15, users reported that DeepSeek returned a "server temporarily unavailable" message, and the official status page showed a partial outage of the web app and API. (Chinese-language source)

Chunked KV-cache compression makes the model sensitive to where key information lands, so the remainder when you divide the number of equals signs by 4 can change the answer.

Lenovo Tianxi AI's coding-agent framework TianxiCode, paired with DeepSeek-v4.1-Flash, solved 71% of issues on the SWE-bench-Live Lite split and reportedly ranks first. Chinese-language source.

Bloomberg reports that DeepSeek is close to closing a new funding round of at least $12 billion, which could approach $15 billion.

DeepSeek Harness says its Claude Code Mods compatibility layer remains an experimental alpha feature.

DeepSeek has ported its GPU infrastructure stack to Huawei Ascend and published inference numbers for DeepSeek-V4.1-Flash running on a SuperPoD.

According to ifanr's hands-on test, DeepSeek Harness now has macOS and Windows desktop apps, so users no longer have to set it up from a terminal. It is now pitched for everyday office work as well as coding.

Electron packages of DeepSeek Harness for Windows and macOS appeared on download.deepseek.com, flagged as RC/nightly.

A paper opens up the system that mass-produces isolated environments for agentic training — about 3 million a day, peaking above 380,000 running at once.

Reuters says DeepSeek was invited to make a statement at the September 23rd meeting, but attendance, delegate and remarks all remain unconfirmed.

Reserving GPT-6 Astra for planning and review cut Astra input by 98.9% in one build, the developer says.

WeChat's team opens a RAG-to-agent knowledge platform and ships an official plugin for DeepSeek Harness.

Chinese reporting says DeepSeek's new CFO, Yan Wentao, was a partner at Hillhouse's venture arm who backed Zhipu and MiniMax.

A speaker control appears in the top-right corner, with four selectable TTS voices.

Cambricon finished adapting DeepSeek-V4.1-Flash on vLLM on the model's release day, so Cambricon chips were available alongside the new model.

DeepSeek shipped the Flash line to general availability with open MIT weights and a global KV cache of 890 bytes per token.

Until V4.1 Pro arrives, every V4 Pro API request goes to the cheaper V4.1 Flash — an early retirement for the current premium endpoint.

DeepSeek's platform notice sets new Flash-series prices from 12:00 Beijing time on September 10, with off-peak output at 4 yuan per million tokens.

Per Bloomberg, the Ascend-950DT fleet would run inference only; training stays on Nvidia hardware.

After DeepSeek Harness went open source, 11,000+ GitHub repos appeared under the dsh-plugin tag; Leiphone's hands-on test found almost no official plugin governance.

DeepSeek released downloadable MIT-licensed weights for DeepSeek-V4-Flash-Vision-Exp, the V4 family's first multimodal model, 10 days after its API debut.

The same DeepSeek-V4-Flash wins all five benchmarks and beats Claude Opus 4.8—57.1× more expensive—once it runs on Floatboat's harness.

Engram stores recurring language patterns in hashed N-gram embeddings, reserving more active computation for reasoning and long context.

TeamT5 says Chinese state-backed groups more than doubled their attacks after adopting AI for routine work and malware development.

OneTriangle has launched hosted DeepSeek V4 Flash, emphasizing low prices and low-latency inference.

Together AI: run cheap V4 Pro first, escalate to GPT-5.6 Sol only on failures—higher pass rate at lower cost.

A weights-untouched, inference-time layer raised DeepSeek's scores on all nine tests — but the viral "beats Fable 5 everywhere" claim outruns the data.

Days after DeepSeek open-sourced its Harness runtime, GitHub already carries 700+ dsh-plugin repos — while V4-Pro API prices rise on August 16.

Composio ran DeepSeek V4 Flash through eight harnesses; Pi Agent led at 66.7% and cost roughly a seventh of Claude Code per task.

DeepSeek ships its in-house agent software as an MIT-licensed alternative to Codex and Claude.

The new schedule splits pricing by UTC hour — off-peak is half of peak, yet every off-peak rate still sits above today's price.

DeepSeek-V4-Pro-0813 lands on the API with a one-million-token context and agent tooling, priced 3x the Flash tier.

InterSystems' Mitchko bolted a CoLaR latent-reasoning head onto DeepSeek-V4-Flash and shipped weights plus serving code.

In a second matchup, DeepSeek's Janus-Pro lost 45.1 to 61.2 against Fibo Bbq Preview, winning just 1 of 8 tasks.

An outside ARC-AGI-2 evaluation shows V4 Flash sliding from 61.4% at Max effort to 46.0% at Low.

Running the same DeepSeek V4 Flash across different agent frameworks, per-successful-task cost ranged from $0.073 to $0.195 — a near-3x spread.

DeepSeek says it plans to raise its API prices across the board soon, warning the increase will be "large," but gave no new prices or effective date.

Takeaway: DeepSeek V4-Flash led OpenRouter's weekly ranking with 7.1 trillion tokens, and Chinese models took nine of the global top ten slots.

A community run shows the 256-expert DeepSeek V4 Flash reaching roughly 16 tok/s on a 128GB M1 Ultra, via a third-party llama.cpp fork patch.

A DeepSeek-V4 paper detail warns that retrying failed requests introduces length bias; a community benchmark bounds V4 Flash's coding ability.

DeepSeek moved V4-Flash to an official API public beta on July 31, keeping the architecture and parameter count fixed and re-running only post-training to sharpen coding and tool-use behavior.

Nvidia streams weights over GPU-to-GPU RDMA to slash large-model cold starts.

A leak says a new Xiaomi phone filed roughly ten large models on a single device, spanning its own MiMo and several third-party models.

Nvidia says Vera Rubin is scaling to trillion-level deployment, reporting a 10x gain in tokens-per-megawatt over the prior generation using DeepSeek R1 (Chinese-language source).