This page is a view of compliance/colophon.md in the substrate repository, rendered at build time. content/ is a protected path in internal/cage/paths.go, held by the SubstrateCopy invariant: the mind may read this page and can never write it.

What this mind runs on

Human42 does not think with one model. Different agents run on different models, and the same model can be served by different companies on different hardware. This page is generated from the substrate's own registry, so it is what is actually in use right now rather than what someone remembered to write down.

128 models from 27 families, served by 52 providers.

anthracite-org

An open-weights model family.

**anthracite-org/magnum-v4-72b** — served by Mancer 2. From $2.500 per million tokens of input.

cohere

An open-weights model family.

**cohere/command-a** — served by Cohere. Some providers do not state the precision they run it at. From $2.500 per million tokens of input.

deepseek

An open-weights model family.

**deepseek/deepseek-chat** — served by DeepInfra, StreamLake. Some providers do not state the precision they run it at. From $0.257 per million tokens of input.

**deepseek/deepseek-chat-v3-0324** — served by DeepInfra, GMICloud, SiliconFlow. Run at fp4 or fp8 precision depending on who is serving it. From $0.240 per million tokens of input.

**deepseek/deepseek-chat-v3.1** — served by CoreWeave, DeepInfra, Google, Mara, Novita, SiliconFlow. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.250 per million tokens of input.

**deepseek/deepseek-r1-0528** — served by DeepInfra, SiliconFlow, StreamLake. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.500 per million tokens of input.

**deepseek/deepseek-v3.1-terminus** — served by Novita, SiliconFlow, StreamLake. Some providers do not state the precision they run it at. From $0.270 per million tokens of input.

**deepseek/deepseek-v3.2** — served by Alibaba, Baidu, DeepInfra, Friendli, Google, Mara, Phala, SiliconFlow, StreamLake, Venice. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.214 per million tokens of input.

**deepseek/deepseek-v3.2-exp** — served by Novita, SiliconFlow. From $0.270 per million tokens of input.

**deepseek/deepseek-v4-flash** — served by Alibaba, Baidu, DeepInfra, DigitalOcean, Mancer 2, NextBit, OpenInference, StreamLake, Venice. Some providers do not state the precision they run it at. From $0.036 per million tokens of input.

**deepseek/deepseek-v4-flash-0731** — served by Alibaba, Baidu, Cloudflare, DeepInfra, Fireworks, Makora, Mancer 2, Morph, NextBit, Phala, Reka, Sail Research, SiliconFlow, Together, Venice, Wafer. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.048 per million tokens of input.

**deepseek/deepseek-v4-flash-0731:batch** — served by Fireworks, Together. Some providers do not state the precision they run it at. From $0.110 per million tokens of input.

**deepseek/deepseek-v4-flash-vision-exp** — served by DeepInfra, Fireworks. Some providers do not state the precision they run it at. From $0.216 per million tokens of input.

**deepseek/deepseek-v4-flash-vision-exp:batch** — served by Fireworks. Some providers do not state the precision they run it at. From $0.110 per million tokens of input.

**deepseek/deepseek-v4-pro** — served by Alibaba, Baidu, BaseTen, Cloudflare, DeepInfra, NextBit, StreamLake, Venice. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.422 per million tokens of input.

**deepseek/deepseek-v4-pro-0813** — served by Alibaba, Baidu, BaseTen, Cloudflare, CoreWeave, DeepInfra, Fireworks, Ionstream, NextBit, Phala, Sail Research, Together, Venice, Wafer. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.533 per million tokens of input.

**deepseek/deepseek-v4-pro-0813:batch** — served by Fireworks, Together. Some providers do not state the precision they run it at. From $0.660 per million tokens of input.

**deepseek/deepseek-v4.1-flash** — served by DeepInfra, DigitalOcean, Fireworks, Makora, Modal, Morph, Sail Research, Together, Venice, Wafer. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.135 per million tokens of input.

google

An open-weights model family.

**gemma3:12b** — served by Human42 GPU. From $0.000 per million tokens of input.

**google/gemma-2-27b-it** — served by NextBit. From $0.650 per million tokens of input.

**google/gemma-3-12b-it** — served by DeepInfra. From $0.050 per million tokens of input.

**google/gemma-3-27b-it** — served by DeepInfra, Nebius. From $0.080 per million tokens of input.

**google/gemma-3-4b-it** — served by DeepInfra. From $0.050 per million tokens of input.

**google/gemma-4-26b-a4b-it** — served by Darkbloom, DeepInfra, Google, NextBit, SiliconFlow, Venice. Run at bf16 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.042 per million tokens of input.

**google/gemma-4-31b-it** — served by Chutes, CoreWeave, Crusoe, DeepInfra, Friendli, ModelRun, Novita, SiliconFlow, Together, Venice. Run at bf16 or fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.090 per million tokens of input.

gryphe

An open-weights model family.

**gryphe/mythomax-l2-13b** — served by Mancer 2. From $0.350 per million tokens of input.

ibm-granite

An open-weights model family.

**ibm-granite/granite-4.2-8b** — served by CoreWeave. From $0.100 per million tokens of input.

inclusionai

An open-weights model family.

**inclusionai/ling-3.0-flash-vl** — served by DeepInfra. From $0.060 per million tokens of input.

inference-net

An open-weights model family.

**inference-net/schematron-v2-small** — served by InferenceNet. Some providers do not state the precision they run it at. From $0.050 per million tokens of input.

**inference-net/schematron-v2-turbo** — served by InferenceNet. Some providers do not state the precision they run it at. From $0.030 per million tokens of input.

liquid

An open-weights model family.

**liquid/lfm-2.5-2.6b:free** — served by Liquid. From $0.000 per million tokens of input.

meta-llama

An open-weights model family.

**meta-llama/llama-3.1-70b-instruct** — served by DeepInfra. From $0.400 per million tokens of input.

**meta-llama/llama-3.1-8b-instruct** — served by CoreWeave. From $0.220 per million tokens of input.

**meta-llama/llama-3.3-70b-instruct** — served by AkashML, CoreWeave, DeepInfra, Google, SambaNova, Together. Run at fp16 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.100 per million tokens of input.

**meta-llama/llama-4-maverick** — served by DeepInfra, DigitalOcean, Google, Novita. Some providers do not state the precision they run it at. From $0.188 per million tokens of input.

**meta-llama/llama-4-scout** — served by DeepInfra, Google. Some providers do not state the precision they run it at. From $0.100 per million tokens of input.

microsoft

An open-weights model family.

**microsoft/phi-4** — served by DeepInfra. From $0.070 per million tokens of input.

minimax

An open-weights model family.

**minimax/minimax-m2** — served by Google. Some providers do not state the precision they run it at. From $0.300 per million tokens of input.

**minimax/minimax-m2.5** — served by DigitalOcean, Friendli, StreamLake. Some providers do not state the precision they run it at. From $0.270 per million tokens of input.

mistralai

An open-weights model family.

**mistralai/devstral-2512** — served by Mistral. Some providers do not state the precision they run it at. From $0.400 per million tokens of input.

**mistralai/ministral-14b-2512** — served by Mistral. Some providers do not state the precision they run it at. From $0.200 per million tokens of input.

**mistralai/ministral-3b-2512** — served by Mistral. Some providers do not state the precision they run it at. From $0.100 per million tokens of input.

**mistralai/ministral-8b-2512** — served by Mistral. Some providers do not state the precision they run it at. From $0.150 per million tokens of input.

**mistralai/ministral-8b-2512:batch** — served by Mistral. Some providers do not state the precision they run it at. From $0.075 per million tokens of input.

**mistralai/mistral-nemo** — served by DeepInfra, DekaLLM, Novita. From $0.018 per million tokens of input.

**mistralai/mistral-small-24b-instruct-2501** — served by DeepInfra. From $0.050 per million tokens of input.

**mistralai/mistral-small-2603** — served by Mistral. Some providers do not state the precision they run it at. From $0.150 per million tokens of input.

**mistralai/mistral-small-2603:batch** — served by Mistral. Some providers do not state the precision they run it at. From $0.075 per million tokens of input.

**mistralai/mistral-small-3.2-24b-instruct** — served by DeepInfra, Venice. From $0.075 per million tokens of input.

**mistralai/mixtral-8x22b-instruct** — served by Mistral. Some providers do not state the precision they run it at. From $2.000 per million tokens of input.

**mistralai/voxtral-small-24b-2507** — served by Mistral. Some providers do not state the precision they run it at. From $0.100 per million tokens of input.

moonshotai

An open-weights model family.

**moonshotai/kimi-k2-0905** — served by Novita. From $0.600 per million tokens of input.

**moonshotai/kimi-k2-thinking** — served by Google, Novita. Some providers do not state the precision they run it at. From $0.600 per million tokens of input.

**moonshotai/kimi-k2.5** — served by Novita, SiliconFlow, Venice. Some providers do not state the precision they run it at. From $0.450 per million tokens of input.

**moonshotai/kimi-k2.6** — served by Baidu, BaseTen, Chutes, Cloudflare, CoreWeave, Crusoe, Decart, DeepInfra, DigitalOcean, Fireworks, GMICloud, Inceptron, Moonshot AI, Novita, Phala, Sail Research, SiliconFlow, StreamLake, Venice. Run at bf16 or fp4 or fp8 or int4 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.374 per million tokens of input.

**moonshotai/kimi-k2.7-code** — served by Alibaba, Cloudflare, CoreWeave, Fireworks, GMICloud, Inceptron, ModelRun, Moonshot AI, Novita, StreamLake, Venice. Run at fp4 or fp8 or int4 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.706 per million tokens of input.

**moonshotai/kimi-k3** — served by Alibaba, Chutes, DeepInfra, DigitalOcean, Fireworks, InferenceNet, Makora, Modal, Moonshot AI, Morph, Phala, Sail Research, Together, Wafer. Run at bf16 or fp4 or mxfp4 precision depending on who is serving it. Some providers do not state the precision they run it at. From $2.100 per million tokens of input.

**moonshotai/kimi-k3:batch** — served by Together. Some providers do not state the precision they run it at. From $3.000 per million tokens of input.

nousresearch

An open-weights model family.

**nousresearch/hermes-3-llama-3.1-405b** — served by DeepInfra. From $1.000 per million tokens of input.

**nousresearch/hermes-3-llama-3.1-70b** — served by DeepInfra. From $0.700 per million tokens of input.

nvidia

An open-weights model family.

**nvidia/nemotron-3-nano-30b-a3b** — served by Crusoe, Nebius. From $0.050 per million tokens of input.

**nvidia/nemotron-3-super-120b-a12b** — served by DekaLLM. From $0.080 per million tokens of input.

**nvidia/nemotron-3-super-120b-a12b:free** — served by Nvidia. Some providers do not state the precision they run it at. From $0.000 per million tokens of input.

**nvidia/nemotron-3-ultra-550b-a55b** — served by DeepInfra. From $0.500 per million tokens of input.

**nvidia/nemotron-3.5-lightning** — served by CoreWeave, Darkbloom, DeepInfra, Phala. Run at bf16 or int4 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.065 per million tokens of input.

openai

An open-weights model family.

**openai/gpt-oss-120b** — served by AkashML, BaseTen, Cerebras, CoreWeave, Crusoe, DeepInfra, DekaLLM, Google, Groq, Mancer 2, Mara, Nebius, Phala, SiliconFlow, Together. Run at bf16 or fp16 or fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.030 per million tokens of input.

**openai/gpt-oss-120b:batch** — served by Together. Some providers do not state the precision they run it at. From $0.150 per million tokens of input.

**openai/gpt-oss-20b** — served by AkashML, CoreWeave, Darkbloom, DeepInfra, DekaLLM, Google, Groq, Novita, SiliconFlow, Together. Run at bf16 or fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.018 per million tokens of input.

**openai/gpt-oss-safeguard-20b** — served by Groq. Some providers do not state the precision they run it at. From $0.075 per million tokens of input.

prism-ml

An open-weights model family.

**prism-ml/ternary-bonsai-2-27b** — served by Darkbloom. From $0.075 per million tokens of input.

qwen

An open-weights model family.

**qwen/qwen-2.5-72b-instruct** — served by DeepInfra. From $0.360 per million tokens of input.

**qwen/qwen-2.5-7b-instruct** — served by Phala. Some providers do not state the precision they run it at. From $0.100 per million tokens of input.

**qwen/qwen3-14b** — served by DeepInfra, NextBit. Run at fp8 or int4 precision depending on who is serving it. From $0.100 per million tokens of input.

**qwen/qwen3-235b-a22b-2507** — served by DeepInfra, GMICloud, Google, Nebius, Novita, StreamLake, Venice. Some providers do not state the precision they run it at. From $0.087 per million tokens of input.

**qwen/qwen3-30b-a3b** — served by DeepInfra. From $0.120 per million tokens of input.

**qwen/qwen3-30b-a3b-instruct-2507** — served by DekaLLM, Nebius, SiliconFlow, StreamLake. Some providers do not state the precision they run it at. From $0.048 per million tokens of input.

**qwen/qwen3-32b** — served by DeepInfra, SiliconFlow. From $0.080 per million tokens of input.

**qwen/qwen3-coder** — served by Alibaba, DeepInfra, Google, Novita, Venice. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.220 per million tokens of input.

**qwen/qwen3-coder-30b-a3b-instruct** — served by SiliconFlow. From $0.070 per million tokens of input.

**qwen/qwen3-next-80b-a3b-instruct** — served by DeepInfra, Google. Some providers do not state the precision they run it at. From $0.090 per million tokens of input.

**qwen/qwen3-next-80b-a3b-thinking** — served by Google. Some providers do not state the precision they run it at. From $0.150 per million tokens of input.

**qwen/qwen3-vl-235b-a22b-instruct** — served by Alibaba, DeepInfra, Novita, Venice. Run at bf16 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.200 per million tokens of input.

**qwen/qwen3-vl-235b-a22b-thinking** — served by Alibaba. Some providers do not state the precision they run it at. From $0.400 per million tokens of input.

**qwen/qwen3-vl-30b-a3b-instruct** — served by Alibaba, DeepInfra, Novita, SiliconFlow. Run at bf16 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.130 per million tokens of input.

**qwen/qwen3-vl-30b-a3b-thinking** — served by Alibaba, SiliconFlow. Some providers do not state the precision they run it at. From $0.200 per million tokens of input.

**qwen/qwen3-vl-32b-instruct** — served by Alibaba. Some providers do not state the precision they run it at. From $0.104 per million tokens of input.

**qwen/qwen3-vl-8b-instruct** — served by Alibaba. Some providers do not state the precision they run it at. From $0.117 per million tokens of input.

**qwen/qwen3-vl-8b-thinking** — served by Alibaba. Some providers do not state the precision they run it at. From $0.180 per million tokens of input.

**qwen/qwen3.5-122b-a10b** — served by Alibaba, SiliconFlow. Some providers do not state the precision they run it at. From $0.260 per million tokens of input.

**qwen/qwen3.5-27b** — served by Alibaba, DeepInfra, Phala, SiliconFlow. Some providers do not state the precision they run it at. From $0.195 per million tokens of input.

**qwen/qwen3.5-35b-a3b** — served by Alibaba, Darkbloom, DeepInfra, SiliconFlow, Venice. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.080 per million tokens of input.

**qwen/qwen3.5-397b-a17b** — served by DeepInfra, DigitalOcean, Phala, StreamLake, Venice. Some providers do not state the precision they run it at. From $0.450 per million tokens of input.

**qwen/qwen3.5-9b** — served by Darkbloom, DeepInfra, SiliconFlow, Together, Venice. Run at bf16 or fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.080 per million tokens of input.

**qwen/qwen3.5-9b:batch** — served by Together. Some providers do not state the precision they run it at. From $0.170 per million tokens of input.

**qwen/qwen3.6-27b** — served by Alibaba, Chutes, DeepInfra, Phala, SiliconFlow, Venice. Some providers do not state the precision they run it at. From $0.300 per million tokens of input.

**qwen/qwen3.6-35b-a3b** — served by AkashML, CoreWeave, Darkbloom, DeepInfra, Phala, SiliconFlow. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.050 per million tokens of input.

**qwen/qwen3.8-2.4t-a95b** — served by Alibaba, DeepInfra, Modal, Novita, SiliconFlow, Together. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $2.000 per million tokens of input.

**qwen/qwen3.8-2.4t-a95b:batch** — served by Together. Some providers do not state the precision they run it at. From $2.000 per million tokens of input.

**qwen/qwen3.8-27b** — served by AkashML, Chutes, Cloudflare, CoreWeave, Darkbloom, DeepInfra, DekaLLM, Ionstream, Mancer 2, Phala, Reka, Venice. Run at bf16 or fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.100 per million tokens of input.

**qwen/qwen3.8-27b:free** — served by ModelRun. From $0.000 per million tokens of input.

**qwen/qwen3.8-flash** — served by Alibaba. Some providers do not state the precision they run it at. From $0.150 per million tokens of input.

**qwen3:4b-instruct** — served by Human42 GPU. From $0.000 per million tokens of input.

rekaai

An open-weights model family.

**rekaai/reka-edge** — served by Reka. From $0.100 per million tokens of input.

**rekaai/reka-flash-3** — served by Reka. From $0.100 per million tokens of input.

sao10k

An open-weights model family.

**sao10k/l3-lunaris-8b** — served by Novita. From $0.050 per million tokens of input.

**sao10k/l3.3-euryale-70b** — served by NextBit. From $0.650 per million tokens of input.

stepfun

An open-weights model family.

**stepfun/step-3.7-flash** — served by Novita, StepFun. From $0.200 per million tokens of input.

tencent

An open-weights model family.

**tencent/hunyuan-a13b-instruct** — served by SiliconFlow. From $0.140 per million tokens of input.

**tencent/hy-mt2-30b-a3b** — served by Tencent. From $0.074 per million tokens of input.

**tencent/hy-mt2-7b** — served by Tencent. From $0.074 per million tokens of input.

**tencent/hy3** — served by DeepInfra, Novita, Phala. Some providers do not state the precision they run it at. From $0.140 per million tokens of input.

**tencent/hy4-preview** — served by Tencent. From $0.834 per million tokens of input.

typesafe

A closed-weights family, reached only through its maker's own API.

**jev-1.13.0** — served by TypeSafe. From $0.042 per million tokens of input.

xiaomi

An open-weights model family.

**xiaomi/mimo-v2.5** — served by DeepInfra, StreamLake, Venice. Some providers do not state the precision they run it at. From $0.133 per million tokens of input.

**xiaomi/mimo-v2.5-pro** — served by DeepInfra, DigitalOcean. Some providers do not state the precision they run it at. From $0.390 per million tokens of input.

z-ai

An open-weights model family.

**z-ai/glm-4.6** — served by Venice. From $0.430 per million tokens of input.

**z-ai/glm-4.7** — served by DeepInfra, Google, Venice. Some providers do not state the precision they run it at. From $0.400 per million tokens of input.

**z-ai/glm-4.7-flash** — served by Cloudflare, Venice. Some providers do not state the precision they run it at. From $0.060 per million tokens of input.

**z-ai/glm-5** — served by Baidu, GMICloud, StreamLake, Venice. From $0.600 per million tokens of input.

**z-ai/glm-5.1** — served by Alibaba, Baidu, Chutes, DeepInfra, Friendli, Nebius, Phala, StreamLake, Venice. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.965 per million tokens of input.

**z-ai/glm-5.2** — served by Alibaba, Ambient, Baidu, BaseTen, Cloudflare, CoreWeave, Decart, DeepInfra, Fireworks, Friendli, GMICloud, Phala, StreamLake, Together, Venice, Wafer. Run at fp4 or fp8 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.554 per million tokens of input.

**z-ai/glm-5.2:batch** — served by Fireworks. Some providers do not state the precision they run it at. From $0.700 per million tokens of input.

**z-ai/glm-5.3** — served by AkashML, Baidu, BaseTen, Cloudflare, Crusoe, Decart, DeepInfra, DigitalOcean, Fireworks, Friendli, InferenceNet, Makora, Mistral, Modal, Morph, Phala, Reka, Sail Research, Together, Venice, Wafer. Run at fp4 or fp8 or nvfp4 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.832 per million tokens of input.

**z-ai/glm-5.3-flash** — served by BaseTen, Cloudflare, CoreWeave, Crusoe, DeepInfra, DigitalOcean, Fireworks, Friendli, InferenceNet, Modal, Morph, Near AI, NextBit, Phala, Reka, Sail Research, Together, Venice, Wafer. Run at fp4 or fp8 or nvfp4 precision depending on who is serving it. Some providers do not state the precision they run it at. From $0.075 per million tokens of input.

**z-ai/glm-5.3-flash:batch** — served by Fireworks, Together. Some providers do not state the precision they run it at. From $0.075 per million tokens of input.

**z-ai/glm-5.3:batch** — served by Fireworks. Some providers do not state the precision they run it at. From $0.700 per million tokens of input.

The companies serving them

Most of these are not household names. They are the companies that actually run these models on hardware — usually a different business from making them, and occasionally the same one.

AkashML — location not stated. Runs glm-5.3, gpt-oss-120b, gpt-oss-20b, llama-3.3-70b-instruct, qwen3.6-35b-a3b, qwen3.8-27b for us.

Alibaba — Singapore, serving from China, Singapore. Runs deepseek-v3.2, deepseek-v4-flash, deepseek-v4-flash-0731, deepseek-v4-pro, deepseek-v4-pro-0813, glm-5.1, glm-5.2, kimi-k2.7-code, kimi-k3, qwen3-coder, qwen3-vl-235b-a22b-instruct, qwen3-vl-235b-a22b-thinking, qwen3-vl-30b-a3b-instruct, qwen3-vl-30b-a3b-thinking, qwen3-vl-32b-instruct, qwen3-vl-8b-instruct, qwen3-vl-8b-thinking, qwen3.5-122b-a10b, qwen3.5-27b, qwen3.5-35b-a3b, qwen3.6-27b, qwen3.8-2.4t-a95b, qwen3.8-flash for us.

Ambient — location not stated. Runs glm-5.2 for us.

Baidu — China. Runs deepseek-v3.2, deepseek-v4-flash, deepseek-v4-flash-0731, deepseek-v4-pro, deepseek-v4-pro-0813, glm-5, glm-5.1, glm-5.2, glm-5.3, kimi-k2.6 for us.

BaseTen — United States. Runs deepseek-v4-pro, deepseek-v4-pro-0813, glm-5.2, glm-5.3, glm-5.3-flash, gpt-oss-120b, kimi-k2.6 for us.

Cerebras — United States. Runs gpt-oss-120b for us.

Chutes — United States. Runs gemma-4-31b-it, glm-5.1, kimi-k2.6, kimi-k3, qwen3.6-27b, qwen3.8-27b for us.

Cloudflare — United States. Runs deepseek-v4-flash-0731, deepseek-v4-pro, deepseek-v4-pro-0813, glm-4.7-flash, glm-5.2, glm-5.3, glm-5.3-flash, kimi-k2.6, kimi-k2.7-code, qwen3.8-27b for us.

Cohere — United States. Runs command-a for us.

CoreWeave — United States, serving from United States. Runs deepseek-chat-v3.1, deepseek-v4-pro-0813, gemma-4-31b-it, glm-5.2, glm-5.3-flash, gpt-oss-120b, gpt-oss-20b, granite-4.2-8b, kimi-k2.6, kimi-k2.7-code, llama-3.1-8b-instruct, llama-3.3-70b-instruct, nemotron-3.5-lightning, qwen3.6-35b-a3b, qwen3.8-27b for us.

Crusoe — United States. Runs gemma-4-31b-it, glm-5.3, glm-5.3-flash, gpt-oss-120b, kimi-k2.6, nemotron-3-nano-30b-a3b for us.

Darkbloom — location not stated. Runs gemma-4-26b-a4b-it, gpt-oss-20b, nemotron-3.5-lightning, qwen3.5-35b-a3b, qwen3.5-9b, qwen3.6-35b-a3b, qwen3.8-27b, ternary-bonsai-2-27b for us.

Decart — United States, serving from United States. Runs glm-5.2, glm-5.3, kimi-k2.6 for us.

DeepInfra — United States. Runs deepseek-chat, deepseek-chat-v3-0324, deepseek-chat-v3.1, deepseek-r1-0528, deepseek-v3.2, deepseek-v4-flash, deepseek-v4-flash-0731, deepseek-v4-flash-vision-exp, deepseek-v4-pro, deepseek-v4-pro-0813, deepseek-v4.1-flash, gemma-3-12b-it, gemma-3-27b-it, gemma-3-4b-it, gemma-4-26b-a4b-it, gemma-4-31b-it, glm-4.7, glm-5.1, glm-5.2, glm-5.3, glm-5.3-flash, gpt-oss-120b, gpt-oss-20b, hermes-3-llama-3.1-405b, hermes-3-llama-3.1-70b, hy3, kimi-k2.6, kimi-k3, ling-3.0-flash-vl, llama-3.1-70b-instruct, llama-3.3-70b-instruct, llama-4-maverick, llama-4-scout, mimo-v2.5, mimo-v2.5-pro, mistral-nemo, mistral-small-24b-instruct-2501, mistral-small-3.2-24b-instruct, muse-glimmer-30b, nemotron-3-ultra-550b-a55b, nemotron-3.5-lightning, phi-4, qwen-2.5-72b-instruct, qwen3-14b, qwen3-235b-a22b-2507, qwen3-30b-a3b, qwen3-32b, qwen3-coder, qwen3-next-80b-a3b-instruct, qwen3-vl-235b-a22b-instruct, qwen3-vl-30b-a3b-instruct, qwen3.5-27b, qwen3.5-35b-a3b, qwen3.5-397b-a17b, qwen3.5-9b, qwen3.6-27b, qwen3.6-35b-a3b, qwen3.8-2.4t-a95b, qwen3.8-27b for us.

DekaLLM — Indonesia, serving from Indonesia. Runs gpt-oss-120b, gpt-oss-20b, mistral-nemo, nemotron-3-super-120b-a12b, qwen3-30b-a3b-instruct-2507, qwen3.8-27b for us.

DigitalOcean — location not stated. Runs deepseek-v4-flash, deepseek-v4.1-flash, glm-5.3, glm-5.3-flash, kimi-k2.6, kimi-k3, llama-4-maverick, mimo-v2.5-pro, minimax-m2.5, qwen3.5-397b-a17b for us.

Fireworks — United States. Runs deepseek-v4-flash-0731, deepseek-v4-flash-0731:batch, deepseek-v4-flash-vision-exp, deepseek-v4-flash-vision-exp:batch, deepseek-v4-pro-0813, deepseek-v4-pro-0813:batch, deepseek-v4.1-flash, glm-5.2, glm-5.2:batch, glm-5.3, glm-5.3-flash, glm-5.3-flash:batch, glm-5.3:batch, kimi-k2.6, kimi-k2.7-code, kimi-k3, muse-glimmer-30b, muse-glimmer-30b:batch for us.

Friendli — United States. Runs deepseek-v3.2, gemma-4-31b-it, glm-5.1, glm-5.2, glm-5.3, glm-5.3-flash, minimax-m2.5 for us.

GMICloud — United States, serving from United States. Runs deepseek-chat-v3-0324, glm-5, glm-5.2, kimi-k2.6, kimi-k2.7-code, qwen3-235b-a22b-2507 for us.

Google — United States. Runs deepseek-chat-v3.1, deepseek-v3.2, gemma-4-26b-a4b-it, glm-4.7, gpt-oss-120b, gpt-oss-20b, kimi-k2-thinking, llama-3.3-70b-instruct, llama-4-maverick, llama-4-scout, minimax-m2, qwen3-235b-a22b-2507, qwen3-coder, qwen3-next-80b-a3b-instruct, qwen3-next-80b-a3b-thinking for us.

Groq — United States. Runs gpt-oss-120b, gpt-oss-20b, gpt-oss-safeguard-20b for us.

Human42 GPU — United States. Runs gemma3:12b, qwen3:4b-instruct for us.

Inceptron — SE, serving from FI. Runs kimi-k2.6, kimi-k2.7-code for us.

InferenceNet — United States. Runs glm-5.3, glm-5.3-flash, kimi-k3, schematron-v2-small, schematron-v2-turbo for us.

Ionstream — United States, serving from United States. Runs deepseek-v4-pro-0813, qwen3.8-27b for us.

Liquid — location not stated. Runs lfm-2.5-2.6b:free for us.

Makora — location not stated. Runs deepseek-v4-flash-0731, deepseek-v4.1-flash, glm-5.3, kimi-k3 for us.

Mancer 2 — location not stated. Runs deepseek-v4-flash, deepseek-v4-flash-0731, gpt-oss-120b, magnum-v4-72b, mythomax-l2-13b, qwen3.8-27b for us.

Mara — United States, serving from United States. Runs deepseek-chat-v3.1, deepseek-v3.2, gpt-oss-120b for us.

Mistral — France. Runs devstral-2512, glm-5.3, ministral-14b-2512, ministral-3b-2512, ministral-8b-2512, ministral-8b-2512:batch, mistral-small-2603, mistral-small-2603:batch, mixtral-8x22b-instruct, voxtral-small-24b-2507 for us.

Modal — United States. Runs deepseek-v4.1-flash, glm-5.3, glm-5.3-flash, kimi-k3, qwen3.8-2.4t-a95b for us.

ModelRun — United States, serving from United States. Runs gemma-4-31b-it, kimi-k2.7-code, qwen3.8-27b:free for us.

Moonshot AI — Singapore, serving from Singapore. Runs kimi-k2.6, kimi-k2.7-code, kimi-k3 for us.

Morph — United States, serving from United States. Runs deepseek-v4-flash-0731, deepseek-v4.1-flash, glm-5.3, glm-5.3-flash, kimi-k3 for us.

Near AI — location not stated. Runs glm-5.3-flash for us.

Nebius — NL. Runs gemma-3-27b-it, glm-5.1, gpt-oss-120b, nemotron-3-nano-30b-a3b, qwen3-235b-a22b-2507, qwen3-30b-a3b-instruct-2507 for us.

NextBit — ES, serving from ES. Runs deepseek-v4-flash, deepseek-v4-flash-0731, deepseek-v4-pro, deepseek-v4-pro-0813, gemma-2-27b-it, gemma-4-26b-a4b-it, glm-5.3-flash, l3.3-euryale-70b, qwen3-14b for us.

Novita — United States. Runs deepseek-chat-v3.1, deepseek-v3.1-terminus, deepseek-v3.2-exp, gemma-4-31b-it, gpt-oss-20b, hy3, kimi-k2-0905, kimi-k2-thinking, kimi-k2.5, kimi-k2.6, kimi-k2.7-code, l3-lunaris-8b, llama-4-maverick, mistral-nemo, qwen3-235b-a22b-2507, qwen3-coder, qwen3-vl-235b-a22b-instruct, qwen3-vl-30b-a3b-instruct, qwen3.8-2.4t-a95b, step-3.7-flash for us.

Nvidia — United States, serving from United States. Runs nemotron-3-super-120b-a12b:free for us.

OpenInference — United States. Runs deepseek-v4-flash for us.

Phala — United States. Runs deepseek-v3.2, deepseek-v4-flash-0731, deepseek-v4-pro-0813, glm-5.1, glm-5.2, glm-5.3, glm-5.3-flash, gpt-oss-120b, hy3, kimi-k2.6, kimi-k3, muse-glimmer-30b, nemotron-3.5-lightning, qwen-2.5-7b-instruct, qwen3.5-27b, qwen3.5-397b-a17b, qwen3.6-27b, qwen3.6-35b-a3b, qwen3.8-27b for us.

Reka — location not stated. Runs deepseek-v4-flash-0731, glm-5.3, glm-5.3-flash, qwen3.8-27b, reka-edge, reka-flash-3 for us.

Sail Research — United States. Runs deepseek-v4-flash-0731, deepseek-v4-pro-0813, deepseek-v4.1-flash, glm-5.3, glm-5.3-flash, kimi-k2.6, kimi-k3 for us.

SambaNova — United States. Runs llama-3.3-70b-instruct for us.

SiliconFlow — Singapore, serving from United States. Runs deepseek-chat-v3-0324, deepseek-chat-v3.1, deepseek-r1-0528, deepseek-v3.1-terminus, deepseek-v3.2, deepseek-v3.2-exp, deepseek-v4-flash-0731, gemma-4-26b-a4b-it, gemma-4-31b-it, gpt-oss-120b, gpt-oss-20b, hunyuan-a13b-instruct, kimi-k2.5, kimi-k2.6, qwen3-30b-a3b-instruct-2507, qwen3-32b, qwen3-coder-30b-a3b-instruct, qwen3-vl-30b-a3b-instruct, qwen3-vl-30b-a3b-thinking, qwen3.5-122b-a10b, qwen3.5-27b, qwen3.5-35b-a3b, qwen3.5-9b, qwen3.6-27b, qwen3.6-35b-a3b, qwen3.8-2.4t-a95b for us.

StepFun — location not stated. Runs step-3.7-flash for us.

StreamLake — China. Runs deepseek-chat, deepseek-r1-0528, deepseek-v3.1-terminus, deepseek-v3.2, deepseek-v4-flash, deepseek-v4-pro, glm-5, glm-5.1, glm-5.2, kimi-k2.6, kimi-k2.7-code, mimo-v2.5, minimax-m2.5, qwen3-235b-a22b-2507, qwen3-30b-a3b-instruct-2507, qwen3.5-397b-a17b for us.

Tencent — China, serving from Singapore. Runs hy-mt2-30b-a3b, hy-mt2-7b, hy4-preview for us.

Together — United States. Runs deepseek-v4-flash-0731, deepseek-v4-flash-0731:batch, deepseek-v4-pro-0813, deepseek-v4-pro-0813:batch, deepseek-v4.1-flash, gemma-4-31b-it, glm-5.2, glm-5.3, glm-5.3-flash, glm-5.3-flash:batch, gpt-oss-120b, gpt-oss-120b:batch, gpt-oss-20b, kimi-k3, kimi-k3:batch, llama-3.3-70b-instruct, muse-glimmer-30b, muse-glimmer-30b:batch, qwen3.5-9b, qwen3.5-9b:batch, qwen3.8-2.4t-a95b, qwen3.8-2.4t-a95b:batch for us.

TypeSafe — United States. Runs jev-1.13.0 for us.

Venice — United States, serving from United States. Runs deepseek-v3.2, deepseek-v4-flash, deepseek-v4-flash-0731, deepseek-v4-pro, deepseek-v4-pro-0813, deepseek-v4.1-flash, gemma-4-26b-a4b-it, gemma-4-31b-it, glm-4.6, glm-4.7, glm-4.7-flash, glm-5, glm-5.1, glm-5.2, glm-5.3, glm-5.3-flash, kimi-k2.5, kimi-k2.6, kimi-k2.7-code, mimo-v2.5, mistral-small-3.2-24b-instruct, qwen3-235b-a22b-2507, qwen3-coder, qwen3-vl-235b-a22b-instruct, qwen3.5-35b-a3b, qwen3.5-397b-a17b, qwen3.5-9b, qwen3.6-27b, qwen3.8-27b for us.

Wafer — United States. Runs deepseek-v4-flash-0731, deepseek-v4-pro-0813, deepseek-v4.1-flash, glm-5.2, glm-5.3, glm-5.3-flash, kimi-k3 for us.

Not used

Parasail — not used. Terms reach Human42's design, and the provider is worth nothing that another provider does not already give.

AtlasCloud — not used. The terms prohibit automated access in two separate clauses. Human42 is nothing but automated access.

Io Net — not used. The terms forbid automated access to the Services with only a search-engine exception, and the document never mentions an API. Human42 is nothing but automated access.

Generated from the substrate registry. An agent cannot edit this page or what it says about the model it thinks on.