AI Models
Supported models, their token limits and their lifecycle status in neokapi. This page uses the committed catalog that also supplies the binary's provider defaults and output-token budgets.
Each provider lists recommended models first, followed by Advanced & specialised models for higher-cost or specialized use cases. Legacy models remain available but have a newer successor. Select any listed model with --model. Models are removed from the catalog when the provider stops serving them.
“In neokapi since” is the date the model id entered the source tree; models present when the catalog was first seeded share that date, and precise dates are recorded going forward. For what these models cost and how they behave under batching, see the batch eval.
Anthropic
| Model | Status | In neokapi since | Lifecycle | Max output | Context |
|---|---|---|---|---|---|
Claude Sonnet 5 claude-sonnet-5alias: sonnet default for anthropic | active | 2026-07-14 | current | 128k | 1000k |
Claude Haiku 4.5 claude-haiku-4-5alias: haiku | active | 2026-07-03 | current | 64k | 200k |
Advanced & specialised (2) — capable but premium, or tuned for other tasks
| Model | Status | In neokapi since | Lifecycle | Max output | Context |
|---|---|---|---|---|---|
Claude Opus 5 claude-opus-5alias: opus | active | 2026-08-27 | current 1M context is standard, with no long-context surcharge. | 128k | 1000k |
Claude Fable 5 claude-fable-5alias: fable | active | 2026-07-14 | current | 128k | 1000k |
Legacy (4) — superseded, still callable
| Model | Status | In neokapi since | Lifecycle | Max output | Context |
|---|---|---|---|---|---|
Claude Opus 4.8 claude-opus-4-8 | superseded | 2026-07-14 | superseded by Claude Opus 5 1M context is standard, with no long-context surcharge. | 128k | 1000k |
Claude Sonnet 4 claude-sonnet-4 | superseded | 2026-04-03 | superseded by Claude Sonnet 5 Reached in production through the Bedrock inference profile eu.anthropic.claude-sonnet-4-6. | 64k | 200k |
Claude Opus 4 claude-opus-4 | superseded | 2026-07-14 | superseded by Claude Opus 4.8 | 32k | 200k |
Claude 3.5 (Sonnet / Haiku) claude-3-5 | superseded | 2026-06-22 | superseded by Claude Haiku 4.5 | 8k | 200k |
OpenAI
| Model | Status | In neokapi since | Lifecycle | Max output | Context |
|---|---|---|---|---|---|
GPT-5.6 gpt-5.6default for openai | active | 2026-07-14 | current The default id is gpt-5.6-sol, which resolves to this family. | 128k | 1050k |
Advanced & specialised (1) — capable but premium, or tuned for other tasks
| Model | Status | In neokapi since | Lifecycle | Max output | Context |
|---|---|---|---|---|---|
OpenAI o3 o3 | active | 2026-06-22 | current Reasoning model; kept for callers who name it explicitly. | 100k | 200k |
Legacy (5) — superseded, still callable
| Model | Status | In neokapi since | Lifecycle | Max output | Context |
|---|---|---|---|---|---|
GPT-5 gpt-5 | superseded | 2026-07-14 | superseded by GPT-5.6 | 128k | 400k |
GPT-4.1 gpt-4.1 | superseded | 2026-07-14 | superseded by GPT-5.6 | 33k | 1048k |
GPT-4o gpt-4odefault for azureopenai | superseded | 2026-04-03 | superseded by GPT-5.6 Superseded on the OpenAI API, but still the Azure OpenAI default, because Azure deployments lag the public API and the default is deliberately conservative. | 16k | 128k |
GPT-4 (legacy family) gpt-4- | superseded | 2026-07-14 | superseded by GPT-5.6 The trailing dash keeps this prefix from matching gpt-4o or gpt-4.1. | 4k | 128k |
GPT-3.5 gpt-3.5 | superseded | 2026-07-14 | superseded by GPT-5.6 | 4k | 16k |
Gemini
| Model | Status | In neokapi since | Lifecycle | Max output | Context |
|---|---|---|---|---|---|
Gemini 3.5 Flash gemini-3.5-flashdefault for gemini | active | 2026-07-14 | current | 66k | 1049k |
Gemini 3.1 Flash-Lite gemini-3.1-flash-lite | active | 2026-07-14 | current | 64k | 1049k |
Advanced & specialised (2) — capable but premium, or tuned for other tasks
| Model | Status | In neokapi since | Lifecycle | Max output | Context |
|---|---|---|---|---|---|
Gemini 3.1 Pro gemini-3.1-pro | active | 2026-07-14 | current Prompts over 200k input tokens are billed at a higher tier. | 64k | 1049k |
Gemini 3 (family default) gemini-3 | active | 2026-04-03 | current Fallback ceilings for any Gemini 3.x id without a more specific entry. | 64k | 1049k |
Legacy (2) — superseded, still callable
| Model | Status | In neokapi since | Lifecycle | Max output | Context |
|---|---|---|---|---|---|
Gemini 3 Flash gemini-3-flash | superseded | 2026-04-03 | superseded by Gemini 3.5 Flash | 64k | 1049k |
Gemini 2 gemini-2 | superseded | 2026-07-14 | superseded by Gemini 3.5 Flash | 8k | 1049k |
Ollama
| Model | Status | In neokapi since | Lifecycle | Max output | Context |
|---|---|---|---|---|---|
Llama 3.2 (3B) llama3.2default for ollama | active | 2026-06-22 | current Runs locally through Ollama; ceilings depend on the pulled quantisation, so none are pinned here. | — | — |
Catalog reviewed 2026-08-27. Maintained in providers/ai/models.json; make check-models flags when a provider’s live model list diverges from it.