Model comparison

By task: cost/speed/quality (1–5)

Task

Scores are generalized (2025): 5 = best. For private data use local models or enterprise API with DPA. Context = how much text the model sees at once. Check current benchmarks — leaders change fast.

Ranking

GPT-4o mini ✓23.0

bulk tasks · ctx 128k · $×1 · ⚡5 · ★3

Gemini 2.5 Flash22.0

long docs · ctx 1M · $×2 · ⚡5 · ★4

Claude Haiku21.0

classification · ctx 200k · $×2 · ⚡5 · ★3

Llama 3.3 (local)19.0

privacy · ctx 128k · $×1 · ⚡3 · ★3

GPT-4o17.0

general · ctx 128k · $×4 · ⚡4 · ★5

Claude Sonnet 417.0

code, analysis · ctx 200k · $×4 · ⚡4 · ★5

Gemini 2.5 Pro15.0

complex tasks · ctx 1M · $×4 · ⚡3 · ★5