OpenCode

2026-10-28

RESEARCH · 1 SOURCE · arXiv cs.AI

kw-bench: benchmark for investigative knowledge work using real VAT reconstructions

kw-bench is a new benchmark (arXiv:2610.36405v1) built from real specialist engagements; its first task is a five-year VAT reconstruction across three generations of SAP delivered as raw system extracts with deterministic grading. The authors ran four harnesses (Claude Code, Codex, OpenCode, micro-cc) on the same model (DeepSeek v4.1 Flash) for two hours — OpenCode 0.542 and micro-cc 0.528, while two lab harnesses detached from their home model scored 0.00; by contrast, micro-cc and Claude Code on Claude Opus 5.5 achieved 0.9986 and 1.000 in 22 and 25 minutes, showing frontier models collapse the task and non-frontier performance depends strongly on the harness.

7.0

2026-10-24

RESEARCH · 1 SOURCE · The Decoder

Study finds agent 'harnesses' drive token costs but rarely improve coding performance

Researchers at Shandong University (Liu and Han) ran five models through three shipped coding-agent harnesses (Claude Code, OpenCode, mini-SWE-agent) on SWE-bench Verified and a larger 447-task pool, and found that swapping harnesses changes results little beyond run-to-run noise while token costs can triple because of large system prompts and repeated context. OpenCode consistently trailed the others on the larger pool (partly due to output limits), and many reported leaderboard gaps fall below statistical detectability given sample sizes and noise.

7.0

2026-10-19

COMPANIES · 1 SOURCE · TechCrunch AI

Hackers breached OpenCode Console, exposed AI API keys and caused unauthorized charges

OpenCode said attackers broke into an older OpenCode Console system, exposing API keys for about 1.4% of its Console users and triggering unauthorized charges on customers' AI accounts; OpenAI, Anthropic and Google blocked affected keys and OpenAI is assisting refunds. OpenCode's open-source tools were not affected, the company gained access to OpenAI's Daybreak Blue models to aid investigation, and the incident — which coincided with the release of GLM-5.3-Flash — highlighted limits where safety filters can hinder defensive analysis.

7.0

2026-10-07

RESEARCH · 1 SOURCE · Microsoft Research

Microsoft Research Asia open-sources Agent Lightning v1.0, a lightweight harnessed agentic RL framework

Microsoft Research Asia released Agent Lightning v1.0, a roughly 3,500-line open-source implementation of a Harnessed Agentic RL training paradigm that trains on the same agent harness used in deployment via an LLM proxy. The framework runs agents as native Kubernetes jobs, avoids reimplementing agent harnesses, and includes an end-to-end coding-agent example that raised Qwen3.5-9B’s Pass@1 on SWE-bench Verified from 41.8% to 56.4% using about 6,000 training samples.

7.0

2026-10-02

COMPANIES · 1 SOURCE · InfoQ AI, ML & Data Engineering

DigitalOcean launches Managed Agents in public preview

DigitalOcean launched DigitalOcean Managed Agents in public preview, a managed cloud infrastructure layer for AI agents that combines a microVM-based Harness Runtime for persistent, isolated agent sessions and an Action Gateway that centralizes and governs tool access (via an MCP endpoint) with connectors to many services; features include pause-when-idle sessions, parallel execution, security-hardened compute/storage, and access to over 16,000 tools.

6.0

2026-09-28

COMPANIES · 1 SOURCE · Hugging Face

Hugging Face Hub adds dedicated RL Environments dataset filter and framework tags

Hugging Face Hub introduced a dedicated spot for reinforcement-learning environments: dataset repositories tagged rl-environment now appear in a new filter and show generated "Use this dataset" commands plus framework compatibility tags (e.g., Harbor, Verifiers, NVIDIA NeMo Gym, OpenEnv). The release focuses on publishing tasksets (tasks, tests, verifiers) in repo datasets to reduce siloed environment formats and to enable running and verifying tasks via Harbor, NeMo Gym, and Hugging Face Jobs/Sandboxes.

6.0

2026-09-23

CODING · 1 SOURCE · AWS Machine Learning

Amazon Bedrock enables OpenCode to run open-weight models for AI coding agents

OpenCode, an open-source terminal-native AI coding agent, can connect to open-weight models via Amazon Bedrock so inference runs securely in a customer’s AWS account without per-seat fees or separate infrastructure. The post demonstrates multi-model workflows (examples: Moonshot AI Kimi K3, OpenAI GPT-OSS 120B, NVIDIA Nemotron 3 Super 120B), and highlights Bedrock features such as regional data-residency profiles, AWS security controls, flexible pricing tiers, and high default throughput limits.

7.0

2026-09-22

MODELS · 1 SOURCE · The Decoder

Xiaomi's MiMo-V2.6-Pro tops open-model rankings; Anthropic alleges Claude data siphoning

Xiaomi released the MiMo‑V2.6 lineup, with flagship MiMo‑V2.6‑Pro (1.02T params, 42B active) scoring 46 on Artificial Analysis's Intelligence Index and claiming very low token costs; the company attributes gains to a major expanded reinforcement‑learning run and published an open RL toolkit, training framework, smaller models, and ~7,000 graded training tasks. Separately, Anthropic's threat report says Xiaomi appears in case GTG‑16008, alleging ~400,000 exchanges in March–April 2026 routed MiMo conversations through OpenClaw/OpenCode to Claude to extract training data as part of a broader set of ~190M exchanges across several labs.

8.0

2026-09-18

MODELS · 1 SOURCE · AWS Machine Learning

Moonshot AI's Kimi K3 (2.8T, 1M-token) now available on Amazon Bedrock

Moonshot AI's Kimi K3 is now available on Amazon Bedrock. Per Moonshot, Kimi K3 is the company's most capable open-weight model (2.8 trillion parameters) with native vision, a 1‑million‑token context window, and ~2.5x scaling-efficiency improvement over Kimi K2; Bedrock also supports explicit prompt caching, Responses/Chat Completions APIs, regional/global inference profiles, and AWS data protections (zero data retention and zero operator access).

7.0

2026-09-16

MODELS · 1 SOURCE · Cohere

Cohere releases North Mini Code — open-source 30B (3B active) agentic coding MoE model

Cohere has open-sourced North Mini Code, a mixture-of-experts agentic coding model sized at 30 billion parameters with 3 billion active, released under an Apache 2.0 license. The model — available on Hugging Face, OpenRouter, and Cohere's Model Vault in bf16, fp8 and w4a16 formats and compatible with OpenCode — is optimized for agentic developer workflows and reports a 33.4 score on the Artificial Analysis Coding Index, higher throughput (up to 2.8x) and ~30% better inter-token latency versus Devstral Small 2 in Cohere's tests.

8.0

2026-09-16

MODELS · 1 SOURCE · Meta AI

Meta launches Muse Spark 1.1, an agentic multimodal model with 1M‑token context

Meta Superintelligence Labs announced Muse Spark 1.1, an upgraded multimodal reasoning model optimized for agentic tasks (tool use, coding, and multimodal workflows). The model supports a 1 million‑token context window, improved multi‑agent orchestration and coding performance, is available in "Thinking" mode in the Meta AI app and on meta.ai, and is accessible to developers via a public preview of the new Meta Model API; Meta says it conducted extensive safety evaluations prior to deployment.

8.0

2026-09-10

MODELS · 1 SOURCE · DeepSeek

DeepSeek launches DeepSeek‑V4.1‑Flash with new causal encoder–decoder and lower costs

DeepSeek announced DeepSeek‑V4.1‑Flash, a smaller-native-vision model using a new Causal Encoder–Decoder design (8B active params for input, 16B for output), new pretraining and larger-scale RL post-training, plus KV‑cache compression. The company is retiring V4‑Flash variants, routing deepseek-v4-pro traffic to V4.1‑Flash from 04:00 UTC on Sept 14, 2026, and changing pricing (new rates take effect 04:00 UTC on Sept 10, 2026) while partners WorkBuddy, CodeBuddy and OpenCode already support V4.1‑Flash.

8.0

2026-04-24

MODELS · 1 SOURCE · DeepSeek

DeepSeek publishes open-source DeepSeek‑V4 preview with 1M-token context

DeepSeek has open-sourced a preview of DeepSeek‑V4, offering two variants: DeepSeek‑V4‑Pro (1.6T total / 49B active params) and DeepSeek‑V4‑Flash (284B total / 13B active params). The company says both models support a default 1M-token context, introduce novel token-wise compression and DSA (DeepSeek Sparse Attention), claim open-source state‑of‑the‑art agentic and reasoning performance (trailing only Gemini‑3.1‑Pro among models they compare to), and are available now via chat.deepseek.com and updated APIs with compatibility for OpenAI ChatCompletions & Anthropic endpoints; older models deepseek-chat and deepseek-reasoner will be retired on Jul 24, 2026 at 15:59 UTC.

8.0