Directory
Endpoints and local runtimes
194 hosted model services and 6 local runtimes: which API each speaks, how it takes a key, where its rate limits, prices and data rules are written, and how to set up OpenCode, Pi, Codex and Claude Code for it.
Hosted services
Services a person reviewed come first, in the order of the review, then the other services by name, then the local runtimes. Payment never changes the order or which services are listed.
A person read each of these services' documentation on the date its page shows.
- OpenAI51 models listed
OpenAI Responses, OpenAI Chat Completions
Structured output: YesTool calling: Yes - Anthropic17 models listed
Anthropic Messages, OpenAI Chat Completions
Structured output: YesTool calling: Yes - Google Gemini API40 models listed
Native API, OpenAI Chat Completions
Structured output: YesTool calling: Yes - Mistral AI39 models listed
OpenAI Chat Completions
Structured output: YesTool calling: Yes - Groq16 models listed
OpenAI Chat Completions, OpenAI Responses
Structured output: YesTool calling: Yes - Cerebras2 models listed
OpenAI Chat Completions
Structured output: YesTool calling: Yes - Together AI32 models listed
OpenAI Chat Completions
Structured output: YesTool calling: Yes - Fireworks AI24 models listed
OpenAI Chat Completions, OpenAI Responses, Anthropic Messages
Structured output: YesTool calling: Yes - DeepInfra76 models listed
OpenAI Chat Completions, Anthropic Messages
Structured output: YesTool calling: Yes - OpenRouter389 models listed
OpenAI Chat Completions, OpenAI Responses, Anthropic Messages
Structured output: YesTool calling: Yes - Ollama Cloud28 models listed
OpenAI Chat Completions, OpenAI Responses, Anthropic Messages, Native API
Structured output: NoTool calling: Yes
183 more services that models.dev lists
These publish an OpenAI-compatible address. Their pages show only what models.dev records.
| Service | Address | Models |
|---|---|---|
| 302.AI | api.302.ai/v1 | 125 |
| Abacus | routellm.abacus.ai/v1 | 108 |
| abliteration.ai | api.abliteration.ai/v1 | 3 |
| above.dev | api.above.dev/v1 | 10 |
| AgentRouter | agentrouter.org/v1 | 5 |
| Agnes AI | apihub.agnes-ai.com/v1 | 4 |
| ai& | api.aiand.com/v1 | 13 |
| AI-ROUTER | api.ai-router.dev/v1 | 5 |
| AI/ML API | api.aimlapi.com/v1 | 177 |
| AI21 Labs | api.ai21.com/studio/v1 | 2 |
| ainetcafe | microquickjs.com/v1 | 1 |
| Aixy | api.aixy-gateway.com/v1 | 13 |
| AKI.IO | aki.io/v1 | 7 |
| Alibaba | dashscope-intl.aliyuncs.com/compatible-mode/v1 | 59 |
| Alibaba (China) | dashscope.aliyuncs.com/compatible-mode/v1 | 91 |
| Alibaba Coding Plan | coding-intl.dashscope.aliyuncs.com/v1 | 10 |
| Alibaba Coding Plan (China) | coding.dashscope.aliyuncs.com/v1 | 10 |
| Alibaba Token Plan | token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1 | 29 |
| Alibaba Token Plan (China) | token-plan.cn-beijing.maas.aliyuncs.com/compatible-mode/v1 | 29 |
| Ambient | api.ambient.xyz/v1 | 17 |
| AMD | developer.amd.com.cn/radeon/api/v1 | 6 |
| AnyAPI | api.anyapi.ai/v1 | 30 |
| Arcee | api.arcee.ai/api/v1 | 7 |
| Auriko | api.auriko.ai/v1 | 15 |
| Bailing | api.tbox.cn/api/llm/v1 | 2 |
| Baseten | inference.baseten.co/v1 | 26 |
| Bee by HEOSSI | api.bee.heossi.com/bee | 6 |
| Berget.AI | api.berget.ai/v1 | 6 |
| Blue Claw | openai.blueclaw.network/v1 | 1 |
| Bothub | openai.bothub.ru/v1 | 8 |
| Boundless | api.inference.boundless.network/v1 | 20 |
| Charm Hyper | hyper.charm.land/v1 | 23 |
| Chutes | llm.chutes.ai/v1 | 14 |
| Clarifai | api.clarifai.com/v2/ext/openai/v1 | 12 |
| Claudinio | api.claudin.io/v1 | 2 |
| ClinePass | api.cline.bot/api/v1 | 18 |
| CloudFerro Sherlock | api-sherlock.cloudferro.com/openai/v1 | 5 |
| Cloudflare Workers AI | api.cloudflare.com/client/v4/accounts/${CLOUDFLARE_ACCOUNT_ID}/ai/v1 | 28 |
| CommandCode | api.commandcode.ai/provider/v1 | 71 |
| CoralBricks | inference.coralbricks.ai/v1 | 1 |
| CoreWeave | api.inference.wandb.ai/v1 | 23 |
| Cortecs | api.cortecs.ai/v1 | 106 |
| CrossModel | api.crossmodel.ai/v1 | 70 |
| Crusoe | api.inference.crusoecloud.com/v1 | 11 |
| D.Run (China) | chat.d.run/v1 | 3 |
| DaoXE | daoxe.com/v1 | 9 |
| DeepSeek | api.deepseek.com | 4 |
| DevPass (LLM Gateway) | api.llmgateway.io/v1 | 216 |
| DigitalOcean | inference.do-ai.run/v1 | 103 |
| DInference | api.dinference.com/v1 | 6 |
| EBCloud | maas-api.ebcloud.com/v1 | 4 |
| Echo | echo.tracerml.ai/v1 | 1 |
| Eden AI | api.edenai.run/v3 | 343 |
| EmpirioLabs AI | api.empiriolabs.ai/v1 | 63 |
| engy | api.engy.ai/v1 | 9 |
| evroc | models.think.evroc.com/v1 | 16 |
| FastRouter | go.fastrouter.ai/api/v1 | 47 |
| Friendli | api.friendli.ai/serverless/v1 | 7 |
| FrogBot | app.frogbot.ai/api/v1 | 26 |
| GitHub Copilot | api.githubcopilot.com | 36 |
| GMI Cloud | api.gmi-serving.com/v1 | 17 |
| GreenPT | api.greenpt.ai/v1 | 40 |
| Helicone | ai-gateway.helicone.ai/v1 | 90 |
| Hetzner | inference.hetzner.com/api/v1 | 2 |
| HPC-AI | api.hpc-ai.com/inference/v1 | 9 |
| Hugging Face | router.huggingface.co/v1 | 82 |
| iFlow | apis.iflow.cn/v1 | 14 |
| Impossibl | api.impossibl.com/v1 | 76 |
| Inception | api.inceptionlabs.ai/v1 | 4 |
| Inceptron | api.inceptron.io/v1 | 4 |
| Inco | api.inco.ai/v1 | 7 |
| Inference | inference.net/v1 | 9 |
| InferX | model.inferx.net/endpoints/v1 | 12 |
| Infomaniak | api.infomaniak.com/2/ai/${INFOMANIAK_PRODUCT_ID}/openai/v1 | 10 |
| IO.NET | api.intelligence.io.solutions/api/v1 | 17 |
| IteraCompute | api.iteracompute.com/v1 | 10 |
| Jalapeno Cloud | api.jalapeno-cloud.ai/v1 | 17 |
| Jiekou.AI | api.jiekou.ai/openai | 61 |
| Kenari | kenari.id/v1 | 64 |
| Kilo Gateway | api.kilo.ai/api/gateway | 397 |
| Kimi For Coding (kimi.ai) | api.kimi.ai/coding/v1 | 4 |
| Kimi For Coding (kimi.com) | api.kimi.com/coding/v1 | 4 |
| klokintegration.se | api-gw.klok.ipaas.se/proxy/kloker-key/v1 | 3 |
| Kosmik Compute | api.koscompute.com/v1 | 1 |
| KUAE Cloud Coding Plan | coding-plan-endpoint.kuaecloud.net/v1 | 1 |
| Lilac | api.getlilac.com/v1 | 4 |
| Llama | api.llama.com/compat/v1 | 7 |
| LLM Gateway | api.llmgateway.io/v1 | 438 |
| LLM Tech | api.llmtech.eu/v1 | 1 |
| LLMTR | llmtr.com/v1 | 32 |
| LongCat | api.longcat.chat/openai | 1 |
| LowRouter | api.lowrouter.ai/v1 | 31 |
| LucidQuery | api.lucidquery.com/v1 | 4 |
| Meganova | api.meganova.ai/v1 | 19 |
| Melious | api.melious.ai/v1 | 13 |
| Mixlayer | models.mixlayer.ai/v1 | 5 |
| Moark | moark.com/v1 | 10 |
| Modal | inference.us-west.modal.direct/v1 | 4 |
| Model Oracle AI | api.modeloracle.com/api/v1 | 15 |
| Modelis | modelishub.com/v1 | 9 |
| ModelScope | api-inference.modelscope.cn/v1 | 7 |
| Moonshot AI | api.moonshot.ai/v1 | 4 |
| Moonshot AI (China) | api.moonshot.cn/v1 | 4 |
| Morph | api.morphllm.com/v1 | 3 |
| NaN | api.nan.builders/v1 | 8 |
| NanoGPT | nano-gpt.com/api/v1 | 617 |
| NEAR AI Cloud | cloud-api.near.ai/v1 | 56 |
| Nebius Token Factory | api.tokenfactory.nebius.com/v1 | 24 |
| Neuralwatt | api.neuralwatt.com/v1 | 31 |
| Nova | api.nova.amazon.com/v1 | 2 |
| Novita AI | api.novita.ai/openai | 86 |
| Nvidia | integrate.api.nvidia.com/v1 | 107 |
| OCI Generative AI | inference.generativeai.us-chicago-1.oci.oraclecloud.com/openai/v1 | 9 |
| Ofox | api.ofox.ai/v1 | 154 |
| OpenCode | opencode.ai/inference/openai/v1 | 128 |
| OpenCode Go | opencode.ai/inference/go/openai/v1 | 33 |
| OpenReason | api.openreason.app/v1 | 3 |
| Opper | api.opper.ai/v3/compat | 57 |
| OrcaRouter | api.orcarouter.ai/v1 | 117 |
| OVHcloud AI Endpoints | oai.endpoints.kepler.ai.cloud.ovh.net/v1 | 12 |
| Pareto Inference | api.paretoinference.com/v1 | 1 |
| Pendra | api.pendra.ai/api/v1 | 6 |
| Pioneer | api.fastino.ai/v1 | 117 |
| Poe | api.poe.com/v1 | 142 |
| Poolside | inference.poolside.ai/v1 | 3 |
| Prime Inference | api.pinference.ai/api/v1 | 68 |
| QiHang | api.qhaigc.net/v1 | 9 |
| Qiniu | api.qnaigc.com/v1 | 91 |
| Reflection | api.reflection.ai/openai/v1 | 1 |
| Regolo AI | api.regolo.ai/v1 | 18 |
| Requesty | router.requesty.ai/v1 | 180 |
| routing.run | api.routing.run/v1 | 15 |
| RunInfra | api.runinfra.ai/v1 | 8 |
| Sakana AI | api.sakana.ai/v1 | 4 |
| SambaNova | api.sambanova.ai/v1 | 6 |
| Sarvam AI | api.sarvam.ai/v1 | 2 |
| Scaleway | api.scaleway.ai/v1 | 18 |
| SCNet Token Plan | api.scnet.cn/api/llm/v1 | 19 |
| SCX.ai | api.scx.ai/v1 | 4 |
| SenseNova (China) | token.sensenova.cn/v1 | 6 |
| SiliconFlow | api.siliconflow.com/v1 | 57 |
| SiliconFlow (China) | api.siliconflow.cn/v1 | 44 |
| STACKIT | api.openai-compat.model-serving.eu01.onstackit.cloud/v1 | 8 |
| Standard Compute | api.stdcmpt.com/v1 | 40 |
| StepFun (China) | api.stepfun.com/v1 | 9 |
| StepFun (Global) | api.stepfun.ai/v1 | 9 |
| StepFun Step Plan (China) | api.stepfun.com/step_plan/v1 | 5 |
| StepFun Step Plan (Global) | api.stepfun.ai/step_plan/v1 | 4 |
| submodel | llm.submodel.ai/v1 | 9 |
| Synthetic | api.synthetic.new/openai/v1 | 11 |
| Tempr Gateway | api.temprhq.io/v1 | 82 |
| Tencent Coding Plan (China) | api.lkeap.cloud.tencent.com/coding/v3 | 8 |
| Tencent Token Plan | api.lkeap.cloud.tencent.com/plan/v3 | 2 |
| Tencent TokenHub | tokenhub.tencentmaas.com/v1 | 3 |
| TensorX | api.tensorx.ai/v1 | 25 |
| The Grid AI | api.thegrid.ai/v1 | 16 |
| Tinfoil | inference.tinfoil.sh/v1 | 8 |
| TokenGo | api.tokengo.com/v1 | 13 |
| TokenRouter | api.tokenrouter.com/v1 | 1 |
| TrustedRouter | api.trustedrouter.com/v1 | 7 |
| Umans AI | api.code.umans.ai/v1 | 7 |
| Umans AI Coding Plan | api.code.umans.ai/v1 | 8 |
| UnoRouter | api.unorouter.com/v1 | 23 |
| Upstage | api.upstage.ai/v1/solar | 5 |
| Vancine | vancine.com/v1 | 8 |
| Vispark | api.lab.vispark.in/v1 | 3 |
| Volcengine Ark | ark.cn-beijing.volces.com/api/v3 | 16 |
| Volcengine Ark Coding Plan | ark.cn-beijing.volces.com/api/coding/v3 | 10 |
| Vultr | api.vultrinference.com/v1 | 15 |
| Wafer | pass.wafer.ai/v1 | 5 |
| Wallaby | api.wallabytoken.com/v1 | 1 |
| Xiaomi | api.xiaomimimo.com/v1 | 9 |
| Xiaomi Token Plan (China) | token-plan-cn.xiaomimimo.com/v1 | 9 |
| Xiaomi Token Plan (Europe) | token-plan-ams.xiaomimimo.com/v1 | 9 |
| Xiaomi Token Plan (Singapore) | token-plan-sgp.xiaomimimo.com/v1 | 9 |
| Xpersona | www.xpersona.co/v1 | 13 |
| Z.AI | api.z.ai/api/paas/v4 | 18 |
| Z.AI Coding Plan | api.z.ai/api/coding/paas/v4 | 7 |
| Zeldoc | api.zeldoc.ai/v1 | 1 |
| Zenifra | ai.zenifra.com/v1 | 1 |
| ZenMux | zenmux.ai/api/v1 | 130 |
| Zhipu AI | open.bigmodel.cn/api/paas/v4 | 17 |
| Zhipu AI Coding Plan | open.bigmodel.cn/api/coding/paas/v4 | 4 |
Local runtimes
Each runs models on your own hardware and answers on your own computer. The start command takes a model from its page.
- Ollamagguf files
OpenAI Chat Completions, OpenAI Responses, Anthropic Messages
ollama run hf.co/{repository}:{quantization} - LM Studiogguf, mlx files
OpenAI Chat Completions, OpenAI Responses, Anthropic Messages
lms get {repository} lms server start - llama.cpp servergguf files
OpenAI Chat Completions, OpenAI Responses, Anthropic Messages
llama-server -hf {repository}:{quantization} --jinja - vLLMsafetensors files
OpenAI Chat Completions, OpenAI Responses, Anthropic Messages
vllm serve {huggingface_id} - SGLangsafetensors files
OpenAI Chat Completions, Anthropic Messages
sglang serve --model-path {huggingface_id} --port 30000 - MLX LMmlx files
OpenAI Chat Completions
mlx_lm.server --model {mlx_repository}
What each harness reads
| Harness | OpenAI Chat Completions | OpenAI Responses | Anthropic Messages | Source |
|---|---|---|---|---|
| OpenCode | Reads it | Reads it | No | Documentation, read 2026-09-24 |
| Pi | Reads it | Reads it | Reads it | Documentation, read 2026-09-24 |
| Codex | No | Reads it | No | Documentation, read 2026-09-24 |
| Claude Code | No | No | Reads it | Documentation, read 2026-09-24 |
Sources and dates
Every row keeps the address of each source it uses and the day it was read. A fact no source states is shown as Unknown. The directory republishes only sources whose licence or terms allow it. It was built on .
| Source | What it gives | Licence or terms | Rows | Last read |
|---|---|---|---|---|
| Hugging Face Hub API | Licence, parameter count, publication date, downloads and task of open models, read within the API's published request limit. | Terms of service; each model card declares its own licence, checked | 2,018 | 2026-10-20 |
| Hugging Face model configurations | Layers, key and value heads, head size and context length, the numbers the memory formula needs. | Terms of service, checked | 825 | 2026-10-20 |
| Hugging Face GGUF file lists | The quantized files of a model and their sizes, from the GGUF copy its maker or the most downloaded copier publishes. | Terms of service, checked | 607 | 2026-10-20 |
| models.dev | Provider prices with each record's own date, the limits and capabilities of the makers' own APIs, release dates, and each provider's documented address and key variable. | MIT, checked | 804 | 2026-10-20 |
| LiteLLM model price and context file | Provider prices as listed on the day the file was read, and the input and output limits and tool calling, structured output and reasoning flags of the makers' own APIs. | MIT, checked | 353 | 2026-10-20 |
| LMArena leaderboard dataset | The overall text arena score with style control from the latest leaderboard, rounded to a whole number, for a hosted model the arena names with its maker's own identifier. | CC BY 4.0, checked | 69 | 2026-10-19 |
| Baltor provider client records | Output limits that Baltor's own provider clients declare or observed, each with its day. | Baltor's own record, checked | 7 | 2026-09-04 |
| Provider and runtime documentation | API styles, authentication, structured output, tool calling, rate-limit, pricing and data-policy pages, read by a person on the day each fact names. | Baltor's own record of each provider's public pages, checked | 17 | 2026-09-27 |
| Harness documentation | How OpenCode, Pi, Codex and Claude Code read a model provider, from each harness's own documentation. | Baltor's own record of each harness's documentation, checked | 0 | Unknown |
Linked, never copied:
- Ollama library: Linked only. Ollama's terms refuse automated access without permission, so no data is copied from it. Terms
- OpenRouter: Linked only. OpenRouter's terms of August 31, 2026 forbid copying information on its site or services by script and using it for a competing service, so none of its prices, limits or rankings are copied here. It documents a Models API that your own tool can read with your own account. Terms
- Artificial Analysis: Linked only. Its website terms allow only personal, noncommercial use, so none of its indexes, prices or speed measurements are copied here. Terms
Paid links
No link in this directory is a paid link or an ad, and no listing is paid for. The order and the contents of every list come from the sources named on this page.