agent harness

7 articles tagged with agent harness

November 11, 2026
product update

Alibaba merges Qoder's agent harness into open-source Qwen Code 1.0 Preview under Apache-2.0

Alibaba announced Qwen Code 1.0 on November 11, merging the agent harness from its Qoder coding product into the open-source Qwen Code project. The preview sits on a separate branch of QwenLM/qwen-code, and Alibaba says developers can now embed the harness in their own products.

November 8, 2026
benchmark

Co-Design Bench: best coding agent fully solves only 26.4% of real user change requests

Researchers from Alibaba and several Asian universities released Co-Design Bench, a coding-agent test built from real user change requests. According to the paper, the best of ten models, GPT-6 Astra, meets about 70 percent of a task's requirements on average but is fully correct on only 26.4 percent of tasks.

November 5, 2026
product updateMoonshot AI

Moonshot AI launches Kimi Hosted Agents, a managed agent runtime with sandboxes at 0.072 yuan per hour

Moonshot AI has launched Kimi Hosted Agents, a Harness as a Service product that runs the agent runtime on Kimi's infrastructure. It offers isolated sandboxes, exportable audit logs, more than 140 skills and more than 30 plugins. IT Home cites a sandbox rate of 0.072 yuan per hour plus model and plugin charges.

November 3, 2026
benchmarkIBM

IBM's CUGA harness passes 81.1% of AppWorld tasks on GLM-5.3-flash; Hermes 46.7%, Deep Agents 30.0%

IBM Research held the model constant (open-weight GLM-5.3-flash) and varied only the agent harness on 90 AppWorld tasks. CUGA, IBM's own open-source harness, passed 81.1% of tasks, against 46.7% for Hermes and 30.0% for Deep Agents.

October 31, 2026
product update

Factory releases Droid 1.0 agent harness, 65.2% on Terminal-Bench 4 with Claude Opus 5.5

Factory released Droid 1.0, a model-independent coding agent harness that makes programmatic tool calling the default. On Factory's own Terminal-Bench 4 runs, Droid scores 65.2% with Claude Opus 5.5 at $4.73 per passed task.

October 4, 2026
research

Google's RRSI cuts overfitting in self-improving agents: up to 4.7-point gains on unseen tasks with ~30% fewer tokens

Google Cloud AI Research and several universities introduced RRSI, a method that stops self-optimizing agent harnesses from memorizing their test tasks. According to the paper, it gains up to 4.7 points on five unseen benchmarks and uses about 30% fewer runtime tokens than the unregularized version, with the underlying model frozen.

October 2, 2026
product updateAnthropic

Pi 1.0 agent harness goes stable; Pi Durable ports it to TypeScript with crash-resumable state

Pi, the minimalist agent harness now under Earendil, has reached version 1.0. A companion release, Pi Durable, ports it to TypeScript and externalizes all stateful components so agents can resume after crashes. Pricing and licensing were not disclosed in the source.