Model catalogue
One API. Pick a model by slug, route via plan, get charged by credits.
Chat
GPT-4o mini
stableim/openai-gpt-4o-mini
Cheap, fast OpenAI chat model. Good general-purpose default.
- Provider
- OpenAI
- Context
- 128,000 tokens
- Input
- $0.2000 / M tok
- Output
- $0.8000 / M tok
- Speed
- 85/100
- Quality
- 70/100
GPT-4o
stableim/openai-gpt-4o
Strong OpenAI multimodal model. Vision, tool use, structured outputs.
- Provider
- OpenAI
- Context
- 128,000 tokens
- Input
- $3.0000 / M tok
- Output
- $12.0000 / M tok
- Speed
- 65/100
- Quality
- 92/100
Claude 3.5 Sonnet
stableim/anthropic-claude-3-5-sonnet
Anthropic flagship. Top coding, reasoning, and vision.
- Provider
- Anthropic
- Context
- 200,000 tokens
- Input
- $4.0000 / M tok
- Output
- $18.0000 / M tok
- Speed
- 60/100
- Quality
- 95/100
Claude 3.5 Haiku
stableim/anthropic-claude-3-5-haiku
Fast cheap Anthropic model. Great for high-volume tasks.
- Provider
- Anthropic
- Context
- 200,000 tokens
- Input
- $1.0000 / M tok
- Output
- $5.0000 / M tok
- Speed
- 88/100
- Quality
- 78/100
Llama 3.1 70B (Turbo)
stableim/together-llama-3-1-70b
Meta Llama 3.1 70B Instruct via Together AI. Strong open-source baseline.
- Provider
- Together AI
- Context
- 131,072 tokens
- Input
- $1.1000 / M tok
- Output
- $1.1000 / M tok
- Speed
- 70/100
- Quality
- 80/100
Hermes 3 (Llama 3.1 70B)
stableim/together-hermes-3-llama-3.1-70b
Nous Research fine-tune of Llama 3.1 70B. Less RLHF-aligned, follows user direction more directly.
- Provider
- Together AI
- Context
- 131,072 tokens
- Input
- $1.1000 / M tok
- Output
- $1.1000 / M tok
- Speed
- 70/100
- Quality
- 84/100
Hermes 3 (Llama 3.1 405B)
stableim/together-hermes-3-llama-3.1-405b
Nous Research's flagship Hermes finetune on Llama 3.1 405B. Top open-source reasoning + steerability.
- Provider
- Together AI
- Context
- 131,072 tokens
- Input
- $4.4000 / M tok
- Output
- $4.4000 / M tok
- Speed
- 45/100
- Quality
- 93/100
DeepSeek V3
stableim/together-deepseek-v3
DeepSeek's 671B MoE flagship. Trained in China — different alignment lens than US labs.
- Provider
- Together AI
- Context
- 131,072 tokens
- Input
- $1.5000 / M tok
- Output
- $1.5000 / M tok
- Speed
- 60/100
- Quality
- 92/100
Qwen 2.5 72B Instruct
stableim/together-qwen-2.5-72b
Alibaba's flagship 72B. Strong at code/math, multilingual, less Western-aligned.
- Provider
- Together AI
- Context
- 32,768 tokens
- Input
- $1.5000 / M tok
- Output
- $1.5000 / M tok
- Speed
- 70/100
- Quality
- 87/100
Llama 3.3 70B
stableim/together-llama-3.3-70b
Meta's newest 70B. Performance close to Llama 3.1 405B at a fraction of the cost.
- Provider
- Together AI
- Context
- 131,072 tokens
- Input
- $1.1000 / M tok
- Output
- $1.1000 / M tok
- Speed
- 72/100
- Quality
- 88/100
Llama 3.1 405B
stableim/together-llama-3.1-405b
Meta's largest open-weights model. Top-tier reasoning and knowledge.
- Provider
- Together AI
- Context
- 131,072 tokens
- Input
- $4.4000 / M tok
- Output
- $4.4000 / M tok
- Speed
- 40/100
- Quality
- 94/100
Mixtral 8x22B
stableim/together-mixtral-8x22b
Mistral's open MoE flagship. Strong code + multilingual, very permissive defaults.
- Provider
- Together AI
- Context
- 65,536 tokens
- Input
- $1.5000 / M tok
- Output
- $1.5000 / M tok
- Speed
- 68/100
- Quality
- 84/100
WizardLM-2 8x22B
stableim/together-wizardlm-2-8x22b
Microsoft Research WizardLM-2 on Mixtral 8x22B. Strong general capability, instruction-tuned.
- Provider
- Together AI
- Context
- 65,536 tokens
- Input
- $1.5000 / M tok
- Output
- $1.5000 / M tok
- Speed
- 65/100
- Quality
- 85/100
Reasoning
DeepSeek R1
stableim/together-deepseek-r1
DeepSeek R1 reasoning model. Shows full chain-of-thought, strong math/code.
- Provider
- Together AI
- Context
- 131,072 tokens
- Input
- $3.6000 / M tok
- Output
- $8.4000 / M tok
- Speed
- 35/100
- Quality
- 96/100
Embeddings
text-embedding-3-small
stableim/openai-text-embedding-3-small
OpenAI embedding model. 1536 dims, cheap, fast.
- Provider
- OpenAI
- Context
- 8,192 tokens
- Input
- $0.0250 / M tok
- Speed
- 95/100
- Quality
- 78/100