Open source models are getting scary good at specialized tasks.
Open models + fine-tuning can beat general models on your specific problem.
What changed
• Fine-tuned open models competitive with closed models on specific tasks
• Hosting costs decreased significantly with optimization
• Community tooling for fine-tuning and deployment matured
Who it affects
• Teams with specialized domains
• Cost-sensitive applications
• Privacy-focused organizations
What to do now
• Benchmark open models on your specific tasks before dismissing them
• Calculate total cost including hosting, not just API costs
• Start with hosted open model APIs before committing to self-hosting
• Join communities around models relevant to your domain
Related updates
GPT-5.2 pushes harder on real work: code, tools, long context.
More useful for shipping work, especially with structure.
Claude Opus 4.5 leans into coding + agents, with stronger robustness.
Better for serious coding + agent work, especially when you wire it properly.
Gemini 3 expands reasoning + multimodal capability across Google products.
Gemini's getting more capable, especially inside Google's own stack.
Context windows are now absurdly long, but retrieval still matters.
Long contexts are impressive, but retrieval isn't obsolete yet.