Kimi K2.7-Code: Open Weights, First-Party Numbers (2026)
June 13, 2026
Moonshot's Kimi K2.7-Code is a 1T open-weight coding model with cheap API pricing, but every launch benchmark is first-party. Here's what's actually verified.
#llm-benchmarks
Claude Fable 5: Anthropic's Mythos-Class Model (2026)
June 10, 2026
Claude Fable 5 is Anthropic's most capable public model yet — a Mythos-class system at $10/$50 that quietly falls back to Claude Opus 4.8 on risky prompts.
Nex-N2-Pro: Open-Weight Coder vs GPT-5.5 (2026)
June 10, 2026
Nex-N2-Pro is a free, open-weight 397B MoE coding model from Nex AGI, built on Qwen3.5. We check its benchmarks against GPT-5.5, Opus, and rival open models.
MiniMax M3: Open-Weight Coding at 1/10 the Cost (2026)
June 9, 2026
MiniMax M3 is an open-weight coding model whose Sparse Attention runs 1M-token context at 1/20th the compute. Its benchmarks beat GPT-5.5 — with caveats.
DeepSeek V4: Open-Weight Frontier at 1/7 the Cost
May 2, 2026
DeepSeek V4 ships 1.6T MoE open weights with 1M-token context: 80.6% on SWE-bench Verified at $1.74/$3.48 per million — roughly 1/7 the output cost of Claude Opus 4.7.
GPT-5.5: OpenAI's First Retrained Base Since GPT-4.5
April 24, 2026
OpenAI released GPT-5.5 on April 23, 2026 — the first fully retrained base since GPT-4.5. Benchmarks, $5/$30 API pricing, 1M context, and Opus 4.7 compared.
Meta Muse Spark: Benchmarks and Strategy (2026)
April 9, 2026
Meta Muse Spark is MSL's first proprietary model, with top health and science benchmarks but coding gaps. Modes, scores, and what developers should know.
GPT-5.4 Beats Humans at Computer Use: What It Means
April 5, 2026
GPT-5.4 scores 75% on OSWorld, surpassing human experts at desktop tasks. What this means for AI agents, enterprise workflows, and the competition in 2026.
GLM-4.7 Deep Dive: 355B MoE, 200K Context, $0.60/M Tokens
March 8, 2026
Zhipu AI's GLM-4.7 explained: 355B MoE architecture, 200K-token context, multimodal inputs, and $0.60 in / $2.20 out per million tokens on Z.ai.