Tool · 2026-05-12
AI model picker
Answer 5 questions and we'll recommend the best models for your use case.
Top 12 matches for your criteria
$0.02/$0.03 per 1M
- 128,000 ctx
- · tools
- · open weights
- · 25 providers
- · score 41
$0.03/$0.14 per 1M
- 131,072 ctx
- · tools
- · json
- · reasoning
- · open weights
- · 23 providers
- · score 41
$0.04/$0.16 per 1M
- 131,072 ctx
- · tools
- · json
- · reasoning
- · open weights
- · 33 providers
- · score 41
$0.05/$0.23 per 1M
- 128,000 ctx
- · tools
- · open weights
- · 22 providers
- · score 41
$0.26/$0.38 per 1M
- 163,840 ctx
- · tools
- · reasoning
- · open weights
- · 31 providers
- · score 41
$0.1/$0.1 per 1M
- 262,144 ctx
- · tools
- · open weights
- · 18 providers
- · score 40
$0.3/$1.2 per 1M
- 204,800 ctx
- · tools
- · reasoning
- · open weights
- · 40 providers
- · score 40
$0.3/$1.2 per 1M
- 204,800 ctx
- · tools
- · reasoning
- · open weights
- · 25 providers
- · score 40
$0.3/$1.2 per 1M
- 204,800 ctx
- · tools
- · reasoning
- · open weights
- · 23 providers
- · score 40
$0.06/$0.4 per 1M
- 200,000 ctx
- · tools
- · reasoning
- · open weights
- · 18 providers
- · score 40
$0.1/$0.1 per 1M
- 262,144 ctx
- · tools
- · reasoning
- · open weights
- · 17 providers
- · score 40
$0.2/$0.7 per 1M
- 131,072 ctx
- · tools
- · reasoning
- · open weights
- · 18 providers
- · score 40
How the picker scores
- Hard filters: required capabilities (tool calling, vision, etc.) eliminate non-matches first.
- Context filter: models below your minimum context are dropped.
- Budget penalty: models exceeding your per-million-token budget are penalized, not eliminated, so you can see the trade-off.
- Use-case bonus: each preset adds points to capabilities that matter for that workflow.
- Tie-breaker: provider availability (more providers = more production-ready).
Frequently asked questions
How does the picker rank models?
It applies your hard filters first (required capabilities, minimum context, budget cap), then scores remaining models on capability fit + price + provider availability. Models that exceed your budget are penalised but still shown so you can see the trade-off.
Are there model recommendations I cannot get from a normal search?
The picker presets weight capabilities specific to each workflow — e.g. 'agent' boosts tool calling and structured output, 'cheap and good' boosts price efficiency at the expense of context. So the same model can rank differently depending on which preset you pick.
Why are open-weight models sometimes ranked low even when they're cheap?
When you self-host, the API price isn't really the cost — your GPU bill is. The picker reflects only the cheapest hosted price, so open-weight models can underperform vs. their true cost-of-ownership for self-hosters.
Can I save my preset and revisit it?
Your settings persist in the URL — bookmark or share the page after configuring filters. We do not store anything server-side.
What's the difference between this and /best/best-llm-api?
Best-of pages are editor-curated rule-driven lists for popular use cases. The picker is the same data but parameterised — useful when no preset matches your exact constraints (e.g. 'must support vision AND have ≥1M context AND cost < $2/1M').
Explore more
Other tools
Pricing comparisons
Best-of model lists
Browse by capability
Prices in USD per 1M tokens. Unknown means the provider does not publish per-token pricing.
Data is sourced from models.dev and normalized for comparison. Prices and capabilities may change. Always verify critical production decisions with the provider's official documentation.