Blog
RAG vs long context: is stuffing the whole corpus cheaper?
A million-token context window didn't kill RAG. Sending 1M tokens per request uses about 200 times the input of retrieving 5,000, caching narrows the gap but never closes it, and accuracy drops in the middle.
Muse Glimmer 30B vs Qwen3.6 27B: two ~30B open models compared
Meta's Muse Glimmer 30B and Alibaba's Qwen3.6 27B are both Apache 2.0, ~30B dense models. Qwen is stronger on coding; Glimmer's output costs half as much and its cache discount makes agent loops cheap.
Hermes vs OpenClaw: two open-source agents, one API key
Hermes and OpenClaw get compared a lot, but they do different jobs. What each one is for, how to connect both to FastMetal, and what to watch when an always-on agent runs on pay-as-you-go.
Qwen3.8 Max vs Claude Opus 5: Opus for hard code, Qwen for volume
Qwen3.8 Max vs Claude Opus 5: what the published benchmarks actually say, why Qwen's agent scores need a discount, and what the same large coding task costs on each in USD.
DeepSeek V4 Pro vs Claude Opus 5: cheap long context vs quality
DeepSeek V4 Pro vs Claude Opus 5: an MIT-licensed 1.6T open model against Anthropic's closed flagship. What each costs in USD, where each is weak, and when to use which.
Claude Opus 5 vs GPT-5.6 Sol: Opus for quality, Sol for cost
Claude Opus 5 vs GPT-5.6 Sol, the two flagships of July 2026: where each is stronger at agentic coding, what each one gets wrong, and what they cost in USD.
Local LLM vs hosted API: small models local, big ones hosted
Run small models on your own machine with Ollama, and send the open models too big for your memory to a hosted API. Both speak the OpenAI API, so the same code covers both.
Kimi K3 vs GPT-5.6 Sol: has open weights caught the flagship?
Kimi K3, a 2.8-trillion-parameter open-weight model, against OpenAI's GPT-5.6 Sol: agentic coding benchmarks, the weak spots of each, and what both cost in USD.
GPT-5.6 Sol vs Claude Fable 5: benchmarks and field reports disagree
GPT-5.6 Sol vs Claude Fable 5: Fable 5 leads the coding benchmarks, developers prefer Sol as a working agent. The weak spots of each and what both cost in USD.
Grok 4.5 vs GPT-5.6 Sol: value for money vs steadiness
Grok 4.5 and GPT-5.6 Sol launched a day apart in July 2026, both aimed at coding agents. Where each is stronger, what each gets wrong, and what they cost in USD.