Providers
The AI providers available in Alvin and what each is good at.
Alvin brings models from multiple providers into one account. This page describes provider families; it does not guarantee that every direct provider connection is enabled. Use the Model Hub to check current availability for your account, or let Alvin route among the available models.
At a glance
OpenAI
GPT-5.5, GPT-5 mini/nano, GPT-4.1, GPT-4o, image, speech
Anthropic
Claude Opus 4.8, Sonnet 4.6, Haiku 4.5
Gemini 3.5 Flash, 3.1 Pro, 2.5 Pro/Flash
xAI
Grok 4.3, Grok 4
Meta
Llama 4 Scout & Maverick, Llama 3.3
Mistral AI
Mistral Large 3, Small 4, Codestral
DeepSeek
DeepSeek V4 Pro, V4 Flash, V3.2
Cohere
Command A
Perplexity
Sonar Deep Research, Sonar Reasoning Pro, Sonar
Fireworks AI
Qwen 3.5, Kimi K2.6, DeepSeek V4, Llama 4
ElevenLabs
Speech and transcription provider; gateway deployments currently disabled
OpenAI
Strengths — strong instruction following, wide capability coverage, fast.
| Model | Best for | Context |
|---|---|---|
gpt-5.5 | Current flagship, hardest reasoning | 128K |
gpt-5-mini | Balanced, cost-effective | 128K |
gpt-5-nano | Fast and cheap | 128K |
gpt-4.1 | Long context with vision | 1M |
gpt-4o-mini | Cheapest general chat | 128K |
Also: gpt-image-2 and gpt-image-1.5 for images, tts-1-hd for speech, whisper-1 and gpt-4o-transcribe for transcription, embedding models in the wider catalog. The public API does not expose an embeddings endpoint; catalog membership is not endpoint availability.
Anthropic
Strengths — best-in-class for software development, long context, careful and detailed answers, extended thinking.
| Model | Best for | Context |
|---|---|---|
claude-opus-4.8 | Current flagship, agentic coding | 200K |
claude-sonnet-4.6 | Coding and agents, everyday driver | 200K |
claude-haiku-4.5 | Fast and affordable | 200K |
Strengths — the largest context windows available, native multimodality, strong price-performance.
| Model | Best for | Context |
|---|---|---|
gemini-3.5-flash | Fast, beats the previous Pro tier on coding | 1M |
gemini-3.1-pro-preview | Most capable, deep reasoning | 1M |
gemini-3.1-flash-lite | Ultra-fast, lowest cost | 1M |
Google image and Veo deployments are not currently served by the configured gateway. Check the Model Hub and Studio for available creative tools.
xAI
Strengths — cost-efficient reasoning at long context.
| Model | Best for | Context |
|---|---|---|
grok-4.6 | High-capability reasoning | 1M |
grok-4.5 | Previous-generation reasoning | 1M |
grok-4.3 | Cost-efficient | 1M |
grok-4.20 | Agentic, multi-step work | 1M |
grok-build-0.1 | Code generation | 1M |
Meta
Strengths — open weights, permissive licensing, good value.
| Model | Best for | Context |
|---|---|---|
llama4-maverick | Complex tasks, vision | 1M |
llama4-scout | Efficient, vision | 320K |
llama-3.3-70b | High quality, well understood | 128K |
Mistral AI
Strengths — European hosting, strong code models, efficient at size.
| Model | Best for | Context |
|---|---|---|
mistral-large-3 | Most capable | 256K |
mistral-small-4 | Multimodal, agentic coding | 256K |
codestral-22b | Code generation | 256K |
DeepSeek
Strengths — reasoning and code at a fraction of frontier pricing.
| Model | Best for | Context |
|---|---|---|
deepseek-v4-pro | Flagship reasoning and coding | 128K |
deepseek-v4-flash | Fast, cost-effective | 128K |
Cohere
Strengths — enterprise RAG, retrieval, and embeddings.
| Model | Best for | Context |
|---|---|---|
command-a | Enterprise RAG | 256K |
Perplexity
Strengths — the web, with citations. Use these when the answer depends on current information.
| Model | Best for | Context |
|---|---|---|
sonar-deep-research | Multi-step research with citations | 128K |
sonar-reasoning-pro | Reasoning grounded in live sources | 128K |
perplexity-sonar | Fast web-grounded answers | 128K |
Fireworks AI
Strengths — fast, cheap serving of open-weight models.
Hosts Qwen 3.5, Kimi K2.6, DeepSeek V4 and Llama 4 among others. Useful when you want an open model with better latency than self-hosting gives you.
ElevenLabs
ElevenLabs provides speech technology, but its direct gateway deployments are currently disabled. Do not assume that an ElevenLabs model can be selected through the public API. See the current model list and Studio for available speech options.
Picking a provider
Most users don't need to. Alvin routing selects among the models available for the request and your account.
If you want to choose manually, the Model Hub lets you filter by capability, context size and cost, and compare candidates side by side before committing.
Related
- Model Hub — browse the catalog
- Comparison — compare models side by side
- Supported models — full list with pricing