#open-source-ai

Z.ai's GLM-5.2 lands with full benchmarks and MIT open weights: #2 on Code Arena Frontend, near Opus 4.8 on agentic coding, at GLM-5.1 pricing.

Cohere's first open-source model is a 30B MoE that runs on a single H100, scores 33.4 on the Coding Index, and ships under Apache 2.0. Full breakdown.

Z.ai shipped GLM-5.2 to every GLM Coding Plan tier with 1M context and long-horizon coding claims. How the distribution-first launch works and how to switch.

H1 2026 open-weight retrospective — DeepSeek V4, Qwen 3, Llama 4 release cadence, benchmark gains vs frontier, adoption patterns, and four trend lines.

MiniMax M2.7 release deep dive — agentic coding benchmarks, tool-use performance, multi-step reasoning, and how it compares to M2.5 and Kimi K2.5.

MiMo V2 Omni complete guide — Xiaomi's omnimodal model with 262K context, unified image/video/audio architecture, and agency deployment patterns.

Complete April 2026 guide to free AI models — Qwen 3.6 Plus, Step 3.5 Flash, NVIDIA Nemotron tiers, and the free-tier options driving OpenRouter.

Complete open-source AI model landscape for April 2026. Gemma 4, Qwen 3.6 Plus, Llama 4, Mistral Small 4, gpt-oss, and GLM-5 ecosystem mapped and compared.

Google Gemma 4 complete guide covering all four variants from 2.3B to 31B parameters. Apache 2.0 license, 128K-256K context, multimodal, Arena #3 open model.

Open-source AI model comparison: Gemma 4 Apache 2.0 vs Llama 4 Meta license vs Mistral Small 4. Benchmarks, licensing, context, and deployment costs.

Zhipu AI releases GLM-5.1 scoring 45.3 on coding benchmarks — 94.6% of Claude Opus 4.6. Huawei-trained, open-source promised. Full analysis and comparison.

Meta FAIR releases TRIBE v2 with 70x resolution improvement for predicting brain activity. Zero-shot capability, open-sourced. Marketing and UX implications.

NVIDIA's Nemotron coalition unites global AI labs to build open frontier models. Membership, model roadmap, training infrastructure, and developer access.

Mistral Small 4 packs 119B parameters (6B active via MoE) with 256K context under Apache 2.0. Unified model for reasoning, vision, and coding.

NVIDIA releases Dynamo 1.0, an open-source distributed inference OS achieving 7x performance boost for Blackwell GPUs. Adopted by major clouds.

Sarvam AI releases 105B and 30B open-source reasoning models under Apache 2.0, trained entirely in India. 128K context window powers the Indus AI assistant.

Compare Meta's Llama 4 Scout and Maverick for business. Benchmarks, deployment costs, fine-tuning guides, and when to choose open-source over proprietary AI.

DeepSeek V4 launches with approximately 1 trillion parameters, 1M context window, and Huawei Ascend optimization. China's frontier multimodal model analysis.

OpenClaw is the fastest-growing AI agent in 2026. Complete setup guide covering installation, skills, messaging integration, and automation workflows.

GLM-5, MiniMax M2.2, Doubao 2.0, Qwen 3.5, and Kimi K2.5 launch before Lunar New Year. Complete guide to China's AI model releases February 2026.

StepFun's Step 3.5 Flash activates only 11B of 196B parameters, delivering 350 tokens/second. Why efficient MoE models are outperforming giants.

Moonshot AI's Kimi K2.5 features 1 trillion parameters and Agent Swarm technology coordinating 100 AI agents. Architecture analysis and use cases.

Deploy open source LLMs in enterprise: GLM-4.7, Mistral Large 3, Qwen 3, MiniMax M2.7. Self-hosting, fine-tuning, compliance, and cost comparison.

Master Devstral 2 (72.2% SWE-bench) and Mistral Vibe CLI. Open-weight coding models that run locally. Complete autonomous agent guide.

Master Mistral 3's 10-model family. Large 3 (675B params), Ministral 3. First open frontier with multimodal + multilingual. Apache 2.0 guide.

Master DeepSeek V3.2 and V3.2-Speciale. IMO gold medal performance, MIT license. 70% cost reduction. Complete open-source AI guide.

Moonshot AI's Kimi K2 Thinking achieves SOTA with 1T parameters, INT4 training, 200-300 tool calls. First open model competitive with GPT-5/Claude.

MiniMax M2 achieves 69.4 on SWE-bench at 8% of Claude's cost. Complete guide to China's open-source AI model for agents, coding & multimodal apps.

Compare GLM 4.6 with Claude Sonnet using current pricing, benchmark caveats, MIT license context, and deployment options.

Qwen3 model-family guide refreshed with 2026 caveats for Qwen3.6-Max-Preview, Coder, Thinking models, and deployment strategies.

DeepSeek-V3.1 guide preserved as historical 2025 context, with 2026 caveats for newer DeepSeek V3.2 and V4 reasoning models.