Articles tagged "オープンソースLLM"

Overview of DeepSeek V4 Preview, released on April 24, 2026: two open-weight Mixture-of-Experts variants (V4-Pro at 1.6T total / 49B active and V4-Flash at 284B / 13B), 1-million-token context, weights on Hugging Face, and rollout via API and chat — based on official information.

Moonshot AI's Kimi K2.6 reached general availability on April 21, 2026. Reported benchmarks: SWE-Bench Pro 58.6 and HLE-Full (with tools) 54.0, leading the open-weight class. Available across Kimi.com, the Kimi app, the API, and Kimi Code CLI; weights on Hugging Face under a Modified MIT License. This is a follow-up update to our earlier comprehensive guide.

Qwen 3.6-27B Dense from Alibaba's Qwen Team, released April 22, 2026: 77.2 on SWE-bench Verified, 59.3 on Terminal-Bench 2.0 (matching Claude 4.5 Opus), 262K-to-1M context, Apache 2.0 license, and 40 tok/s on an RTX 3090 with Q4_K_M — summarized from official sources.

Kimi K2.6, released by Moonshot AI on April 20, 2026, achieves open-source SOTA with HLE w/tools 54.0 and SWE-Bench Pro 58.6, surpassing GPT-5.4 and Claude Opus 4.6. Complete guide covering 13-hour long-horizon coding, 300 parallel agent swarms, and OpenClaw integration.

GLM-5.1 by Z.ai (released April 7, 2026) is the first open-source LLM to top SWE-bench Pro at 58.4%, surpassing GPT-5.4 (57.7%) and Claude Opus 4.6 (57.3%). This guide covers its 744B/40B-active MoE architecture, MIT license, 8-hour autonomous task capability, and setup via Ollama.

MiniMax M2.5 achieves 80.2% on SWE-bench Verified using proprietary Lightning Attention in a 230B MoE model. Full breakdown of architecture, benchmarks, license terms, and setup instructions.

Explore AI adoption strategies for SMBs leveraging Rakuten AI 3.0's Apache 2.0 license and Hugging Face release, achieving up to 90% cost reduction. Learn practical methods to utilize Japanese-specialized strengths in business operations and Rakuten AI Gateway integration possibilities.