AI coding assistants advance as Anthropic merges 46% of Claude Code's pull requests, DeepSeek launches open source rival

AI coding assistants are advancing rapidly as Anthropic and DeepSeek expand their agent capabilities. Anthropic is testing Claude Code on its own software, successfully merging 180 of 388 AI-generated pull requests for a 46% merge rate. Meanwhile, DeepSeek has launched DeepSeek Harness v0.1, an open-source, modular rival to Claude Code, alongside the official release of its agent-optimized V4-Pro model. Concurrently, DeepSeek is raising its API rates by up to 1,100% through a new peak and off-peak pricing structure.

DeepSeek Harness open-source launch

• ▪DeepSeek Harness is built on Cordis, a framework designed around composable plugins, allowing developers to swap out models, tools, sandboxes, filesystems, and user interfaces.
• ▪On its launch day of August 13, 2026, the DeepSeek Harness GitHub repository attracted approximately 27,500 stars and 2,000 forks.
• ▪DeepSeek launched DeepSeek Harness v0.1, an open-source agent harness under the MIT license, as an alternative to integrated coding-agent environments like Anthropic's Claude Code.

DeepSeek V4-Pro agent capabilities

• ▪DeepSeek-V4-Pro-0813 scored 87.9 on Terminal Bench 2.1, 74.1 on Toolathlon-Verified, 71.1 on DSBench-FullStack, and 67.2 on DSBench-Hard in company-reported benchmarks.
• ▪DeepSeek released the official version of DeepSeek-V4-Pro-0813, transitioning the 1.6-trillion-parameter model from preview to general availability with enhanced agent capabilities.
• ▪DeepSeek introduced three configurable reasoning effort levels for the V4 model family: Non-think for routine tasks, Think High for complex planning, and Think Max for difficult problems.

Claude Code autonomous maintenance

• ▪Claude Code created 388 pull requests across Anthropic's repositories over a few weeks, with 180 of those requests merged after human and automated reviews, representing a 46 percent merge rate.
• ▪Anthropic engineer Boris Cherny reported that Claude Code has been running daily maintenance routines on Anthropic's in-house software platforms via a dedicated Slack channel.
• ▪Claude Code runs specialized maintenance routines including a Crash Fuzzer to fix app crashes, a Dup Unifier to merge similar abstractions, and a Dead-Code Remover to strip unused code.

DeepSeek API pricing changes

• ▪Under DeepSeek's new pricing, V4-Pro off-peak rates rise to $0.66 per million cache-miss input tokens and $1.98 output, while peak rates reach $1.32 input and $3.96 output.
• ▪DeepSeek's V4-Pro cache-hit pricing will increase from $0.003625 per million tokens to $0.022 off-peak and $0.044 during peak hours.
• ▪Beginning August 16, 2026, DeepSeek will replace its flat API pricing with peak and off-peak rates, resulting in price increases ranging from 50% to over 1,100% depending on the model and token category.

AI coding agent competitive landscape

• ▪DeepSeek's release of Harness and V4-Pro shifts its competition with Western AI labs like Anthropic and OpenAI from model intelligence and token pricing to the agent execution and workflow layer.
• ▪DeepSeek Harness serves as an open-source, model-agnostic alternative to the agent infrastructure of Claude Code and OpenAI's Codex, supporting file editing, shell execution, and subagent delegation.

2 sources