DeepSeek Harness launches as open source rival to Claude Code, alongside V4-Pro on API with higher prices
Published: August 13, 2026 — VentureBeat (9:47 am PT)
DeepSeek is expanding beyond the model layer and deeper into the software developers use to put AI agents to work.
The Chinese AI lab on Thursday launched the official version of DeepSeek-V4-Pro, an updated flagship model focused heavily on agentic workloads, alongside DeepSeek Harness v0.1, a new open-source agent harness that gives developers an alternative to integrated coding-agent environments such as Anthropic’s Claude Code.
Together, the releases amount to a broader developer push from DeepSeek. V4-Pro is now available across DeepSeek’s web interface, mobile app and API, with native support for the OpenAI Responses API and integration with Codex.
DeepSeek Harness is entering developer preview under the MIT license; code available on GitHub. Built around composable plugins on Cordis framework: “Everything is a plugin” — models, tools, skills, sessions, sandboxes, filesystems, loops, orchestration and UIs. Launch via npx @deepseek-ai/dsh web. Repository warns THERE WILL BE COMPATIBILITY-BREAKING CHANGES.
Harness can inspect repositories, edit files, execute shell commands, search files and the web, maintain plans, invoke skills, delegate to subagents and enforce approval policies. Standard mode is a full coding agent. Local web UI for workspace selection and sensitive-operation approval. Model-agnostic (DeepSeek, Anthropic, OpenAI, custom endpoints). GitHub snapshot Aug 13: ~27,500 stars, ~2,000 forks.
V4-Pro agent-focused upgrade
DeepSeek-V4-Pro-0813 is the official GA of the V4-Pro family (originally previewed April). V4-Pro: 1.6T parameters, 49B activated per token; V4-Flash: 284B / 13B activated; both up to 1M context. Model ID deepseek-v4-pro resolves to latest. Native OpenAI Responses API support; optimized for Codex with one-click setup. Reasoning effort levels: Non-think, Think High, Think Max.
Company-reported agent benchmarks for V4-Pro-0813: 87.9 Terminal Bench 2.1, 74.1 Toolathlon-Verified, 71.1 DSBench-FullStack, 67.2 DSBench-Hard. Public Code Agent tasks tested using upcoming DeepSeek Harness in “minimal mode.”
API pricing reverse
Beginning Aug. 16 at 16:00 UTC, DeepSeek abandons flat API pricing for peak/off-peak rates. Peak hours: 01:00–04:00 UTC and 06:00–10:00 UTC; all other hours off-peak (half of peak).
Current vs new (illustrative):
- V4-Flash: currently 0.28 (cache-miss in / out per 1M) → off-peak 0.66; peak 1.32
- V4-Pro: currently 0.87 → off-peak 1.98; peak 3.96
Reuters: depending on model, token category and time of use, increases ranging from 50% to more than 1,100% over existing rates. Off-peak is 50% cheaper than new peak — not a discount from today’s prices.
For 1M cache-miss in + 1M out: V4-Pro from 2.64 off-peak (~2×) or $5.28 peak (>4×).