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
/
 Qwen2-1.5B-InstructAlibabaApache 2.0 | $0.20/1Mvia AIHubMix | — | 1.5B | $0.20/1M | $0.20/1M | 37K | |
| Qwen3 1.7BAlibabaApache 2.0 | $0.46/1Mvia AIHubMix | — | 2B | $0.046/1M | $0.46/1M | 33K | |
| Qwen3.5 9BAlibabaApache 2.0Prompt caching+4 | $0.13/1Mvia EmpirioLabs AI | 262K | 9.7B | $0.09/1M | $0.13/1M | 19K | |
| Qwen3 8BAlibabaApache 2.0 | $0.06/1Mvia SiliconFlow | — | 8.2B | $0.06/1M | $0.06/1M | 12K | |
 GLM OCRZhipu AIMITVision | $0.028/1Mvia AIHubMix | 32K | 1.3B | $0.03/1M | $0.03/1M | 10K | |
 Gemma 4 E2B ITGoogleApache 2.0 | $0.08/1Mvia Amazon Bedrock | — | 5.1B | $0.04/1M | $0.08/1M | 10K | |
| Qwen3-4BAlibabaApache 2.0 | $0.03/1Mvia Hugging Face | — | 4B | $0.01/1M | $0.03/1M | 10K | |
| Qwen2-7B-InstructAlibabaApache 2.0 | $0.08/1Mvia AIHubMix | — | 7.6B | $0.08/1M | $0.08/1M | 9K | |
 Mistral 7B InstructMistral | $0.059/1Mvia Jiekou.AI | — | 7B | $0.029/1M | $0.059/1M | 8K | |
 Llama 3.2 11B InstructMeta | $0.16/1Mvia Amazon Bedrock | — | 11B | $0.16/1M | $0.16/1M | 7K | |
| Qwen3.5 4BAlibabaApache 2.0Prompt caching+3 | $0.07/1Mvia EmpirioLabs AI | 262K | 4.7B | $0.04/1M | $0.07/1M | 7K | |
| Qwen3-VL-8B-InstructAlibabaApache 2.0Prompt caching+3 | $0.278/1Mvia SiliconFlow | 262K | 8.8B | $0.069/1M | $0.278/1M | 6K | |
| Qwen3-VL-8B-ThinkingAlibabaApache 2.0JSON+3 | $0.694/1Mvia SiliconFlow | 131K | 8.8B | $0.069/1M | $0.694/1M | 6K | |
| Qwen2.5-Coder-7BAlibabaApache 2.0 | $0.03/1Mvia Hugging Face | — | 7.6B | $0.01/1M | $0.03/1M | 6K | |
| Llama-2-13b-chat-hfMetaLlama 2 | $0.50/1Mvia Replicate | — | 13B | $0.10/1M | $0.50/1M | 5K | |
| Gemma 3 4B ITGoogleGemmaPrompt caching+2 | $0.05/1Mvia OpenRouter | 131K | 4.3B | $0.04/1M | $0.05/1M | 4K | |
| Qwen3 0.6BAlibabaApache 2.0 | $0.46/1Mvia AIHubMix | — | 752M | $0.046/1M | $0.46/1M | 4K | |
| Qwen3.5 0.8BAlibabaApache 2.0 | $0.08/1Mvia Prime Inference | — | 873M | $0.04/1M | $0.08/1M | 3K | |
| Qwen3 14BAlibabaApache 2.0JSON+2 | $0.20/1Mvia Hugging Face | 41K | 15B | $0.07/1M | $0.20/1M | 3K | |
| Qwen3.5 2BAlibabaApache 2.0JSON+3 | $0.10/1Mvia Requesty | 262K | 2.3B | $0.02/1M | $0.10/1M | 3K | |
| Ministral 3 3B 2512Mistral | $0.10/1Mvia Mistral | 262K | 3B | $0.10/1M | $0.10/1M | 3K | |
| Qwen2.5-7B-InstructAlibabaApache 2.0JSON+1 | $0.05/1Mvia SiliconFlow | 33K | 7.6B | $0.05/1M | $0.05/1M | 3K | |
| Llama 3.1 8B InstructMetaLlama 3.1JSON+1 | $0.05/1Mvia Helicone | 131K | 8B | $0.02/1M | $0.05/1M | 3K | |
 DeepSeek R1 Distill Qwen 7BDeepSeekMIT | $0.01/1Mvia AIHubMix | — | 7.6B | $0.01/1M | $0.01/1M | 3K | |
| Gemma 3 1B ItGoogleGemma | $0.20/1Mvia AIHubMix | — | 1000M | $0.20/1M | $0.20/1M | 3K | |
| Qwen2.5 3B InstructAlibabaNon-commercial | $0.80/1Mvia AIHubMix | — | 3.1B | $0.40/1M | $0.80/1M | 2K | |
| Gemma 3 12B ITGoogleGemmaJSON+2 | $0.10/1Mvia Helicone | 131K | 12B | $0.05/1M | $0.10/1M | 2K | |
| Ministral 3BMistralPrompt cachingJSON+2 | $0.10/1Mvia DevPass (LLM Gateway) | 131K | 3B | $0.10/1M | $0.10/1M | 2K | |
| LFM2.5-2.6B (free)Liquid AIOpen weights · with termsJSON+2 | $0.50/1Mvia OpenRouter | 66K | 2.7B | $0.35/1M | $0.50/1M | 2K | |
| Gemma 4 12B InstructGoogleApache 2.0 | $0.30/1Mvia SiliconFlow | — | 12B | $0.10/1M | $0.30/1M | 2K | |
| Gemma 2BGoogleGemma | Free/1Mvia Nvidia | — | 2.5B | Free/1M | Free/1M | 1K | |
| Ministral 3 8B 2512Mistral | $0.15/1Mvia Mistral | 262K | 8B | $0.15/1M | $0.15/1M | 1K | |
| Ministral 3 14B 2512Mistral | $0.20/1Mvia Mistral | 262K | 14B | $0.20/1M | $0.20/1M | 1K | |
| Ministral 8BMistralPrompt cachingJSON+2 | $0.143/1Mvia AI/ML API | 262K | 8B | $0.143/1M | $0.143/1M | 340 | |
| Ministral 14B (Arkanecloud)MistralPrompt cachingJSON+2 | $0.20/1Mvia Vercel AI Gateway | 262K | 14B | $0.20/1M | $0.20/1M | 270 | |
 MiniCPM V 4.5OpenBMBApache 2.0 | $1.10/1Mvia Cortecs | — | 8.7B | $0.654/1M | $1.10/1M | 260 | |
 Ahm Phi 3 Medium 128KMicrosoftMIT | $0.68/1Mvia Azure | — | 14B | $0.17/1M | $0.68/1M | — | |
| Ahm Phi 3 Small 128KMicrosoftMIT | $0.60/1Mvia Azure | — | 7.4B | $0.15/1M | $0.60/1M | — | |
| Ahm Phi 3.5 Mini InstructMicrosoftMIT | $3/1Mvia AIHubMix | — | 3.8B | $1/1M | $3/1M | — | |
| Ahm Phi 3.5 Vision InstructMicrosoftMITVision | $1.60/1Mvia AIHubMix | — | 4.1B | $0.40/1M | $1.60/1M | — | |
| Aihubmix Phi 4 Mini (reasoning)MicrosoftMIT | $0.12/1Mvia AIHubMix | 128K | 3.8B | $0.12/1M | $0.12/1M | — | |
 Apertus 8B InstructSwiss Ai | $0.20/1Mvia Hugging Face | — | 8B | $0.10/1M | $0.20/1M | — | |
| AutoGLM Phone 9B MultilingualZhipu AIMIT | $0.138/1Mvia Hugging Face | — | 0.9M | $0.035/1M | $0.138/1M | — | |
| Codegemma 1.1 7BGoogle | Free/1Mvia Nvidia | — | 7B | Free/1M | Free/1M | — | |
| Codegemma 7BGoogleGemma | Free/1Mvia Nvidia | — | 8.5B | Free/1M | Free/1M | — | |
 Command R7BCohereCC BY-NC 4.0JSON | $0.15/1Mvia OpenRouter | 128K | 8B | $0.037/1M | $0.15/1M | — | |
 Cosmos Reason2 8BNVIDIAOpen weights · with terms | Free/1Mvia Nvidia | — | 8.8B | Free/1M | Free/1M | — | |
 Decider Flash V1.2 4BPerplexityApache 2.0Vision | Free/1Mvia Vercel AI Gateway | 262K | 4.5B | $0.01/1M | Free/1M | — | |
| DeepSeek Coder 6.7B InstructDeepSeekOpen weights · with terms | Free/1Mvia Nvidia | — | 6.7B | Free/1M | Free/1M | — | |
| DeepSeek OCR 2DeepSeekApache 2.0 | $0.03/1Mvia Jiekou.AI | — | 3.4B | $0.03/1M | $0.03/1M | — |
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