#DeepSeek

AITinker adds GLM-5.3-Flash and DeepSeek-v4.1-Flash, both of which natively accept image inputs and use efficient attention architecture. GLM-5.3-Flash costs 4-5 times less on Tinker than GLM-5.3. Long-context options for Qwen3.5-4B and Qwen3.6-35B-A3B are also live.

AIByteDance Seed researchers found that language models compressing their KV cache in fixed-size chunks retrieve the same information unevenly depending on token position. In a 128K-token needle-in-a-haystack test, base DeepSeek-V4 checkpoints differed by up to 40.2 percentage points by phase, and post-training narrowed but did not eliminate the gaps. The authors urge evaluating such models across positional phases, since high average accuracy can hide systematic failures.

AIGuizang (@op7418) says he posted news of Anthropic's price cut for Haiku 5.5 and was attacked by commenters who view the pricing as aimed at Chinese models. He compares Haiku 5.5 with DeepSeek-V4.1 flash and Zhipu's GLM 5.3 flash, finding it cheaper than both and scoring one point higher than GLM 5.3 flash on Terminal Bench 4.0 in AA's test.

AIInferact and the vLLM community reported a 1.9× low-concurrency speedup and about 5.3× throughput under a 150 TPS constraint for DeepSeek-V4.1-Flash over three weeks. Gains came from SWA bounded replay with CUDA graphs, which cut TTFT by about 30%, and from integrated DeepSeek kernels such as MegaAttention, Mega-mHC, Mega-Gate, and DeepSelect. The post measures these results on the SemiAnalysis AgentX benchmark.

Why it matters: The post breaks down how SWA bounded replay and fused kernels cut prefill and decode costs, a reusable engineering pattern for long-context agentic serving.

AIDeepSeek is close to raising at least 80 billion yuan ($12 billion), up from an original target of about 50 billion yuan, according to Bloomberg citing people familiar with the matter. Tencent and battery maker CATL are among the largest investors in the round, which is expected to close soon. DeepSeek is planning an IPO in early 2027, though details could still change.

AIARC Prize has published full DeepSeek V4.1 Flash results on its public leaderboard, alongside the benchmark code and testing policy needed to reproduce them. The post links to the complete results page for the model. No benchmark scores are stated in the post itself.

AIFireworks AI has made DeepSeek V4.1 Flash available for training on its Dedicated Training API and Managed Training surfaces. The post positions the model as a strong base for agentic coding, terminal automation, and tool use, and notes it is cost-efficient to serve.

AIHuawei chairman Eric Xu said at Huawei Connect that Ascend now leads Nvidia in China, based on Huawei's own data, but did not give a market share. Bernstein forecasts about 50% for Huawei and 8% for Nvidia this year, and Xu says mainland process nodes, not chip design, are the bottleneck. DeepSeek reportedly plans to deploy at least 160,000 Ascend 950DT chips in Inner Mongolia.

AIDeepSeek released packaged desktop versions of DeepSeek Harness for macOS and Windows. Linux users can install it from the @deepseek-ai/dsh package on npm.

AICline says DeepSeek V4 Pro is now available through ClinePass, its subscription offering roughly 5x discounted access. The model can be used in Cline's new Desktop app, CLI, VS Code, and JetBrains.

AIDeepSeek releases the DeepSeek Harness v0.2 preview with a desktop app for macOS and Windows. The release adds a plugin manager for installing, disabling, and uninstalling plugins without terminal commands, plus an experimental creator mode that generates plugins from user descriptions. The company says DeepSeek Harness is now the most widely used coding agent among users of the official DeepSeek API by DAU and daily sessions.

AIDeepSeek released open weights for V4.1-Flash, which Baseten now offers through its Model APIs. The model has 552B total parameters, 8B active for prefill and 16B for decode, a 1M token context window, and text plus image input. Its Causal Encoder-Decoder design runs only the encoder during prefill and reuses a projected KV cache, and the source reports the global KV cache at a quarter of V4-Flash's memory.

Why it matters: The post explains how the CED architecture splits prefill and decode compute and cuts KV cache memory, which matters for coding agent costs.

AIDeepSeek says V4.1-Flash is now live on its API with native multimodal support, accessed through the model name deepseek-flash. The older V4-Flash and V4-Flash-Vision-Exp are retired, while deepseek-v4-flash and deepseek-v4-flash-vision-exp temporarily route to V4.1-Flash. Requests to deepseek-v4-pro will route to V4.1-Flash at V4.1-Flash rates starting 04:00 UTC on Sept 14, 2026, until V4.1-Pro launches.

AIDeepSeek officially released DeepSeek-V4.1-Flash, the smallest model in its new architecture family, with native multimodal visual understanding. The API now serves it under the model name deepseek-flash, while V4 Flash and V4 Flash Vision Exp were retired and routed to V4.1 Flash. API prices were reduced with the release, and V4 Pro remains available after September 14, 2026.

Why it matters: The release lists benchmark results alongside API model-name changes and retirements, so developers can check both capability claims and migration steps.

AIDeepSeek released DeepSeek-V4.1-Flash, a multimodal Mixture-of-Experts model with 552B backbone parameters and support for contexts up to one million tokens. The technical report says its global KV cache footprint is 890 bytes per token, roughly one quarter of DeepSeek-V4-Flash, and reports 8B activated parameters per token during prefill and 16B during decode.

Why it matters: The report shows KV cache per token falling to about one quarter of DeepSeek-V4-Flash, a concrete tradeoff between long-context serving cost and benchmark results.