Small language models
120 language models with up to 15B parameters: cheap to call, quick to answer, small enough to run on your own hardware.
Makers
/
 DeepSeek R1 0528 Qwen3 8BDeepSeekMIT | $0.09/1Mvia Novita AI | — | 8.2B | $0.06/1M | $0.09/1M | — | |
| DeepSeek R1 Distill Llama 8BDeepSeekMIT | $0.01/1Mvia AIHubMix | — | 8B | $0.01/1M | $0.01/1M | — | |
| DeepSeek R1 Distill Qianfan Llama 8BDeepSeek | $0.548/1Mvia AIHubMix | — | 8B | $0.137/1M | $0.548/1M | — | |
| DeepSeek R1 Distill Qwen 1.5BDeepSeekMIT | $0.01/1Mvia AIHubMix | — | 1.8B | $0.01/1M | $0.01/1M | — | |
| DeepSeek R1 Distill Qwen 14BDeepSeekMIT | $0.10/1Mvia AIHubMix | — | 15B | $0.10/1M | $0.10/1M | — | |
| DeepSeek-OCRDeepSeekMITVision | $0.02/1Mvia AIHubMix | 8K | 3.3B | $0.02/1M | $0.02/1M | — | |
 DeplotGoogleApache 2.0 | Free/1Mvia Nvidia | — | 282M | Free/1M | Free/1M | — | |
 ERNIE 4.5 0.3bBaiduJSONTool calling+1 | $0.054/1Mvia AIHubMix | — | 300M | $0.014/1M | $0.054/1M | — | |
| Gemma 2 9B ItGoogleGemma | $0.03/1Mvia Helicone | — | 9.2B | $0.01/1M | $0.03/1M | — | |
| Gemma 3n E4B ItGoogleGemma | $0.20/1Mvia AIHubMix | — | 7.8B | $0.20/1M | $0.20/1M | — | |
| Gemma 7B ItGoogleGemma | $0.10/1Mvia AIHubMix | — | 8.5B | $0.10/1M | $0.10/1M | — | |
 GLM 4 9B 0414Zhipu AIMIT | $0.05/1Mvia AIHubMix | — | 9.4B | $0.05/1M | $0.05/1M | — | |
| GLM Z1 9B 0414Zhipu AIMIT | $0.05/1Mvia AIHubMix | — | 9.4B | $0.05/1M | $0.05/1M | — | |
| GLM-4.6V-FlashZhipu AIMITPrompt caching+2 | $0.218/1Mvia ZenMux | 200K | 10B | $0.022/1M | $0.218/1M | — | |
| Granite 3.0 3B A800m InstructIBMApache 2.0 | Free/1Mvia Nvidia | — | 3.4B | Free/1M | Free/1M | — | |
| Granite 3.0 8B InstructIBMApache 2.0 | $0.25/1Mvia Replicate | — | 8.2B | $0.05/1M | $0.25/1M | — | |
| Granite 4.0 H MicroIBMApache 2.0JSON | $0.112/1Mvia Cloudflare Workers AI | 131K | 3.2B | $0.017/1M | $0.112/1M | — | |
| Granite 4.2 3BIBMApache 2.0 | $0.12/1Mvia Hugging Face | — | 3.7B | $0.03/1M | $0.12/1M | — | |
| Granite 4.2 8BIBMApache 2.0Prompt caching+3 | $0.25/1Mvia Deep Infra | 131K | 8.8B | $0.06/1M | $0.25/1M | — | |
| Granite 8B Code InstructIBM | Free/1Mvia Nvidia | — | 8B | Free/1M | Free/1M | — | |
| Hermes 2 Pro Llama 3 8BNous ResearchLlama 3 | $0.14/1Mvia Helicone | — | 8B | $0.14/1M | $0.14/1M | — | |
 Hunyuan Mt 7BTencentOpen weights | $0.20/1Mvia AIHubMix | — | 8B | $0.20/1M | $0.20/1M | — | |
| Hy-MT2-1.8BTencentApache 2.0 | $0.177/1Mvia OpenRouter | 8K | 2B | $0.044/1M | $0.177/1M | — | |
| Hy-MT2-7BTencentApache 2.0JSON | $0.295/1Mvia OpenRouter | 8K | 8B | $0.074/1M | $0.295/1M | — | |
| SL3-8B-Lunaris-v1Sao10K | $0.05/1Mvia Deep Infra | 8K | 8B | $0.04/1M | $0.05/1M | — | |
 Llama 2 7BMetaLlama 2 | $0.10/1Mvia AIHubMix | — | 7B | $0.10/1M | $0.10/1M | — | |
| SLlama 3 8B LunarisSao10KJSON | $0.05/1Mvia OpenRouter | 8K | 8B | $0.04/1M | $0.05/1M | — | |
| Llama 3.2 11B Vision InstructMetaLlama 3.2 | $0.20/1Mvia AIHubMix | — | 11B | $0.20/1M | $0.20/1M | — | |
| Llama 3.2 1B InstructMetaLlama 3.2 | $0.10/1Mvia Amazon Bedrock | 60K | 1.2B | $0.10/1M | $0.10/1M | — | |
| Llama 3.2 3B InstructMetaLlama 3.2JSON | $0.05/1Mvia DevPass (LLM Gateway) | 131K | 3.2B | $0.03/1M | $0.05/1M | — | |
| Llama3 8B InstructMetaLlama 3 | $0.25/1Mvia Replicate | — | 8B | $0.05/1M | $0.25/1M | — | |
| Llama3 Groq 8B 8192 Tool Use PreviewMeta | $0.00019/1Mvia AIHubMix | — | 8B | $0.00019/1M | $0.00019/1M | — | |
| Meta Llama 3 8BMetaLlama 3 | $0.25/1Mvia Replicate | — | 8B | $0.05/1M | $0.25/1M | — | |
 Ministral 14BMistralPrompt cachingJSON+2 | $0.20/1Mvia DevPass (LLM Gateway) | 262K | 14B | $0.20/1M | $0.20/1M | — | |
| Ministral 3 14BMistralApache 2.0 | $0.20/1Mvia Amazon Bedrock | — | 14B | $0.20/1M | $0.20/1M | — | |
| Mistral 7B Instruct V0.2MistralApache 2.0 | $0.221/1Mvia Cortecs | — | 7.2B | $0.16/1M | $0.221/1M | — | |
| Mistral Nemo 12B InstructMistral | Free/1Mvia Nvidia | — | 12B | Free/1M | Free/1M | — | |
 Mistral Nemo Minitron 8B 8K InstructNVIDIA | Free/1Mvia Nvidia | — | 8B | Free/1M | Free/1M | — | |
| Mistral-7B-Instruct-v0.1MistralApache 2.0 | $0.19/1Mvia Cloudflare Workers AI | — | 7.2B | $0.11/1M | $0.19/1M | — | |
| Mistral-7B-Instruct-v0.3MistralApache 2.0 | Free/1Mvia Nvidia | — | 7.2B | Free/1M | Free/1M | — | |
| GMythoMax L2 13BGrypheJSON | $0.09/1Mvia Jiekou.AI | 8K | 13B | $0.09/1M | $0.09/1M | — | |
| Nemotron-nano 12b v2-vlNVIDIAReasoningTool calling+1 | $0.60/1Mvia Vercel AI Gateway | 131K | 12B | $0.20/1M | $0.60/1M | — | |
| NVIDIA Nemotron Nano 9B v2NVIDIAOpen weights · with termsReasoning+1 | $0.23/1Mvia Vercel AI Gateway | 131K | 8.9B | $0.06/1M | $0.23/1M | — | |
| NVIDIA-Nemotron-Nano-9B-v2NVIDIAOpen weights · with terms | $0.176/1Mvia AIHubMix | — | 8.9B | $0.044/1M | $0.176/1M | — | |
 PaddleOCR VLPaddlepaddleApache 2.0 | $0.02/1Mvia Novita AI | — | 959M | $0.02/1M | $0.02/1M | — | |
 Palmyra Vision 7BWriter | $0.60/1Mvia Amazon Bedrock | — | 7B | $0.15/1M | $0.60/1M | — | |
 Phi 3 Mini 128K InstructMicrosoftMIT | $0.52/1Mvia Azure | — | 3.8B | $0.13/1M | $0.52/1M | — | |
| Phi 3 Vision 128K InstructMicrosoftMIT | Free/1Mvia Nvidia | — | 4.1B | Free/1M | Free/1M | — | |
| Phi 4MicrosoftMITJSON | $0.14/1Mvia Deep Infra | 16K | 15B | $0.07/1M | $0.14/1M | — | |
| Phi 4 MiniMicrosoftMIT | $0.30/1Mvia Azure | 128K | 3.8B | $0.075/1M | $0.30/1M | — |
Frequently asked questions
There is no fixed line; this page lists models with up to 15 billion parameters, the size of Qwen3 8B or Mistral 7B. Sizes come from each model's official repository or the size in its name.
Most of these are open-weight: a 7–8B model runs on a recent laptop or a single consumer GPU once quantized, and the smallest (1–4B) run on phones. Larger context windows need more memory.
For classification, extraction, summaries, routing and simple chat, often yes, at a fraction of the cost and latency. For hard reasoning or broad knowledge a larger model still wins; compare their scores before deciding.
Can't find what you're looking for? Contact us