DeepSeek V4 and GitHub workflow updates #29

TaeyoungPark

May 27, 2026 — 2 min read

Today's Letter

DeepSeek, V4 Preview and 1M Context Released
GitHub, pull request code coverage preview

DeepSeek, V4 Preview and 1M Context Released

DeepSeek announced DeepSeek-V4 Preview on April 24, 2026, with API availability and open-weight releases on the same day
The release includes DeepSeek-V4-Pro with 1.6T total parameters and 49B active parameters, and DeepSeek-V4-Flash with 284B total and 13B active parameters
DeepSeek says both models support a 1M-token context window by default across its official services, with Thinking and Non-Thinking modes
The API keeps the same base URL and only requires switching the model name to `deepseek-v4-pro` or `deepseek-v4-flash`
DeepSeek also says the models support both OpenAI Chat Completions and Anthropic API compatibility layers
V4-Pro is positioned for stronger reasoning, coding, and agentic workloads, while V4-Flash is presented as the faster and lower-cost option
The company attributes the long-context efficiency to token-wise compression and DeepSeek Sparse Attention, or DSA
DeepSeek says `deepseek-chat` and `deepseek-reasoner` will be retired on July 24, 2026 at 15:59 UTC, and are currently routed to V4-Flash modes

Source: api-docs.deepseek.com
More: github.com · engadget.com · infoworld.com

GitHub, pull request code coverage preview

GitHub put pull request code coverage metrics into public preview for GitHub Code Quality users on github.com
Pull requests can now show an aggregate percentage of code covered, giving reviewers a direct signal on test completeness before merge
Repositories need code coverage enabled and must upload a Cobertura report from the existing CI workflow through the upload-code-coverage action
GitHub Apps and Actions workflows now require the new `code-quality:write` fine-grained permission to upload coverage reports
GitHub says the feature is available for GitHub Enterprise Cloud and Team
GitHub Enterprise Server is not yet supported
The preview is free during the public preview period
The change was announced in the GitHub Changelog on 2026-05-26

Source: github.blog

Jocoletter curates AI, software, and product trends for developers and builders.

#DeepSeek #GitHub

Subscribe to Jocoletter

Anthropic results and NVIDIA training guide #55

Today's Letter 1. Anthropic, Project Fetch Phase Two results published 2. NVIDIA, Low-Precision Transformer Training Guide Anthropic, Project Fetch Phase Two results published * Anthropic published Phase Two results for Project Fetch on June 18, 2026, testing Claude Opus 4.7 on robodog setup and autonomy tasks first run

Agent Infrastructure and Copilot Metrics #54

Today's Letter 1. GitHub adds per-user AI credit metrics to Copilot API 2. AWS, Web Search for Bedrock AgentCore GA GitHub adds per-user AI credit metrics to Copilot API * GitHub added an `ai_credits_used` field to the Copilot usage metrics API for per-user AI credit consumption tracking

Agent deployment, Copilot model shifts, enterprise controls #53

Today's Letter 1. Cloudflare, temporary accounts for agent deployments 2. GitHub Copilot, Opus 4.6 (fast) retirement set for June 29 3. AWS, SageMaker inference metrics on CloudWatch 4. OpenAI adds ChatGPT Enterprise usage analytics and spend controls Cloudflare, temporary accounts for agent deployments * Cloudflare introduced Temporary Cloudflare

Agent Stack, Copilot, and Async Inference #52

Today's Letter 1. GitHub Copilot, context handling and Auto routing update 2. AWS, SageMaker Async Inference adds inline payloads 3. Vercel, Agent Stack and eve unveiled 4. Hugging Face, agentic tooling benchmark for open models published GitHub Copilot, context handling and Auto routing update * GitHub published a June