DeepSeek V4 Pro

Headlines · 3

1. 2d
   DeepSeek services down nearly 5 hours; 'DeepSeekHarness' account surfaces

   Starting at 08:11 Beijing time on October 15, DeepSeek's web, app and API services were partially disrupted, with the official status page showing recovery at 13:04 the same day — a total of 4 hours 52 minutes. Affected services included DeepSeek chat, the DeepSeek V4 Pro API and the DeepSeek V4.1 Flash API (one report instead named a V4.1 Pro API). Third-party monitor StatusGator said it received 86 related reports within 24 hours, and users in Luxembourg, Poland and New York also reported no response or very slow service. Around the same time a Bilibili account named 'DeepSeekHarness' appeared, prompting speculation about a new model or promotion of DeepSeek's open-source agent runtime, though no official evidence supports that link.

   IT之家 AI · 🔥 23
2. 3d
   OpenRouter launches Private Endpoints for fine-tuned models and reserved capacity

   OpenRouter introduced Private Endpoints, letting organizations route their own fine-tuned models, dedicated deployments or reserved capacity through the same API they already use for public models. Private endpoints are visible only to approved organizations and never appear in public model lists, rankings, search, charts or benchmarks. An admin registers a deployment from the account page, and OpenRouter first sends one real request through a saved BYOK key to confirm the key works, the deployment speaks the OpenAI-compatible format, streaming and usage reporting function, and the model ID matches; only then can the endpoint be activated. The full lifecycle — registration, testing, activation, pricing updates, pausing and deletion — is also available through a Management API and a Terraform provider. It is an add-on to the Enterprise plan that must be enabled by an account representative, and each organization can have up to 10 private endpoints by default.

   OpenRouter Blog · 🔥 0
3. Oct 6
   Mistral launches Mistral Large 4 preview, a 1T-parameter open-weight model

   Mistral AI has opened a public preview of Mistral Large 4 (nicknamed Le Chonk), a 1-trillion-parameter natively multimodal open-weight model with roughly 49 billion active parameters; the weights are due by the end of the month. Mistral claims state-of-the-art open-model performance on enterprise workloads such as cybersecurity, finance and law, and says it beats even closed frontier models on visual grounding — though it notes that some cyber-test leads come from Claude and GPT-6 refusing the tasks. Independent measurement is more sobering: Artificial Analysis's Intelligence Index puts ML4 at 38 points, a big jump from Mistral Large 3's 9, but well behind leader Claude Opus 5.5 (Max) at 58.

   Mistral AI News · 🔥 0

Experience and discussion from the community

Share my DeepSeek V4 Pro experienceAsk about DeepSeek V4 Pro