DeepSeek

Independent measurements after DSpark put DeepSeek V4-Pro around 90 tok/s and cut one run from 214s to 116s. The gain matters because it lowers serving cost, though tuning details and memory overhead are still unclear.

OpenRouter said four open-weight models now handle real agentic workloads, and a JPMorgan report put Chinese models at about 45% of platform traffic. The shift matters because teams are optimizing for price, hosting, and task fit instead of defaulting to frontier APIs.

DeepSeek open-sourced DeepSpec, a codebase for training and evaluating draft models for speculative decoding, alongside the DSpark decoding module for V4 checkpoints. It matters because inference teams get a new open stack for improving draft-model quality and decode throughput beyond earlier MTP-style baselines.

Together AI said its DeepSeek V4 Pro deployment now leads Artificial Analysis on both output speed and latency. The claim matters because it turns V4 serving into an inference-systems story about KV cache reuse, prefix reuse, kernels, and endpoint profiles rather than model weights alone.

Warp Agent now accepts user-supplied OpenAI, Anthropic, and Gemini keys plus OpenAI-compatible endpoints such as OpenRouter and DeepSeek. The change removes the paid-plan requirement for inference access and gives terminal users more routing options.

DeepSeek made the temporary 75% V4 Pro discount permanent, cutting first-party pricing to $0.435 per million input tokens and $0.87 output. Artificial Analysis now places it on the cost-performance frontier, but practitioners still question per-task efficiency on harder coding work.

SGLang v0.5.12 added native DeepSeek V4 support with ShadowRadix prefix caching, HiSparse CPU-extended KV, MegaMoE kernels, and Blackwell MLA work. The release broadens hardware targets and improves long-context serving efficiency for open runtimes.

Developers posted new local-model measurements for DS4, Qwen 3.6, and Gemma 4: about 40 tok/s on an M3 Ultra, 70+ tok/s on MacBooks with MPS, and 120-200 tok/s for Qwen3.6-27B on a single RTX 3090. The numbers suggest coding-capable local runs are moving from demos toward regular use.

The vLLM team shipped more than 10 DeepSeek V4 fixes as developers kept posting V4 Pro and Flash results from coding harnesses and local servers. Use the update if serving bugs, cache behavior, or tool-call reliability are blocking cheaper long-context agent runs.

Users reported moving long coding sessions from Claude to DeepSeek V4 Flash and seeing tens of millions of tokens cost only cents. Hacker News discussion also leaned toward Flash over Pro for day-to-day use, so teams should test whether the low published prices hold in their own workflows.

DeepSeek briefly published a paper and threads on point-and-bbox reasoning, about 90 KV entries per 800² image, and RL-trained vision experts, then removed the repo and related mentions. The technique looked like a low-token path to computer use and multimodal reasoning in V4-Flash, but availability and reproducibility are now unclear.

DeepSeek began rolling out Vision beta as a new image-understanding mode in Chat, and early testers reported fast OCR and strong object recognition. The rollout appears limited or staggered, so watch for broader access and formal docs before relying on it.

vLLM 0.20.0 shipped a new CUDA 13 / PyTorch 2.11 / Transformers v5 baseline, TurboQuant 2-bit KV cache, FA4 MLA defaults, and deeper DeepSeek V4 support. The release changes serving baselines across NVIDIA, AMD, Intel, and ARM-CUDA setups, including 4x KV capacity and a clearer upgrade path for teams already running V4.

Independent guides showed DeepSeek V4 running inside Claude Cowork and Claude Code via Anthropic-compatible endpoints, and Ollama added launch commands for Claude-style wrappers. The workflow matters because teams can keep Claude-centered agent UX while sharply lowering model spend, with provider compatibility and setup still the main caveats.

DeepSeek said cache-hit pricing across its API series is now one-tenth of launch levels, on top of the temporary V4-Pro discount through May 5. The cut lowers costs for cache-heavy long-context and agent workloads, so teams should recheck spend assumptions.

SGLang and Miles published a technical breakdown of their DeepSeek V4 day-zero stack, including ShadowRadix caching, Flash Compressor, FP4 expert-weight handling, and measured B200/H200 throughput. That gives deployers concrete serving and training-path numbers for V4 beyond generic launch-day compatibility claims.

DeepSeek lowered V4-Pro API pricing and updated integration guidance for Claude Code, OpenCode, and OpenClaw a day after V4 launched. Check whether V4-Flash is the easier deploy today, while Pro stays heavier and more rate-limited.

OpenClaw shipped a release that routes realtime voice queries to the full agent, defaults new users to V4 Flash, and adds coordinate clicks plus stale-lock recovery for browser automation. It also fixes Telegram, Slack, MCP session, and TTS issues, so update if those flows matter to your setup.

Within a day of launch, vLLM, SGLang, Ollama cloud, OpenCode, Venice, Together, and Baseten added support or hosted access for DeepSeek V4. That makes Flash and Pro easier to test across local, routed, and managed agent stacks.

Engineers unpacked DeepSeek V4's hybrid CSA/HCA attention a day after launch; it claims 27% of V3.2 FLOPs and 10% of its KV cache at 1M tokens. External tests pushed V4 Pro near the top of open-model indexes, but users also reported rate limits and mixed third-party results.

DeepSeek published Tile Kernels, an open-source TileLang repo covering Engram, mHC, MoE routing, and FP4/FP8 kernels, with claims that some are already used in internal training and inference. That matters because it exposes reusable low-level performance work behind DeepSeek’s stack instead of keeping the kernels fully private.

DeepSeek open-sourced V4-Pro and V4-Flash under MIT, with 1M context and aggressive Flash pricing. Day-one support in SGLang, vLLM, and OpenRouter pushes open-weight agentic coding closer to closed frontier models.