Models
Search and compare to find the model you need.
Filters
All140+155+
By Epoch index. Unscored models are filtered out.
All128K1M
Free110+
Showing 285 of 285
- Autocelon/auto
Easy questions go to a light model, hard ones to the top model. If a provider stalls, it reroutes in 0.1s.
Constituent modelsClaude Fable 5 ·GPT-5.2 ·Llama 4 Maverick ·Qwen3 235B ·DeepSeek V3.2 ·GLM-5.2 ·+1 - Devcelon/code
Hard code goes to the top model; lighter code goes to strong, low-cost coding models.
Constituent modelsClaude Fable 5 ·GPT-5.2 ·Qwen3 235B ·Kimi K2.6 ·GLM-5.2 ·DeepSeek V3.2 ·+1 - Top qualitycelon/quality
Always uses top-tier (frontier) models, regardless of difficulty. Use it when answer quality is non-negotiable.
Constituent modelsClaude Fable 5 ·GPT-5.2 ·Claude Opus 4.8 - Top speedcelon/nitro
Routes to the fastest of a quality-vetted pool by measured throughput and latency. For live chat and autocomplete.
Candidate pool281·order decided live by measured speed - Lowest pricecelon/floor
Routes to the cheapest of a quality-vetted pool by live pricing. For bulk processing, classification, and summaries.
Candidate pool281·order decided live by price