Tiny (under 4B) AI models

All 156 open tiny models. Edge / on-device — runs almost anywhere. Compare calculated local requirements, provider-published rented-GPU rates, and dated API rates or clearly labeled estimates.

• Qwen3 0.6B — 800M, Qwen · $0.11/1M API est.
• Qwen2.5 3B Instruct — 3.1B, Qwen · $0.12/1M API est.
• Llama 3.2 3B Instruct — 3B, Meta · $0.19/1M API
• Qwen2.5 1.5B Instruct — 1.5B, Qwen · $0.11/1M API est.
• gemma 3 270m — 300M, google · $0.10/1M API est.
• Qwen3 1.7B — 2B, Qwen · $0.12/1M API est.
• BGE-M3 — 567M, BAAI · $0.10/1M API est.
• Qwen2.5 0.5B Instruct — 500M, Qwen · $0.10/1M API est.
• Qwen2 1.5B Instruct — 1.5B, Qwen · $0.11/1M API est.
• Llama 3.2 1B Instruct — 1.2B, Meta · $0.11/1M API
• Llama 3.2 1B — 1.2B, meta-llama · $0.11/1M API est.
• Qwen2.5 0.5B — 500M, Qwen · $0.10/1M API est.
• Phi-3.5-mini Instruct — 3.8B, Microsoft · $0.13/1M API est.
• gemma 3 1b it — 1B, google · $0.11/1M API est.
• TinyLlama 1.1B Chat v1.0 — 1.1B, TinyLlama · $0.11/1M API est.
• gpt2 large — 800M, openai-community · $0.11/1M API est.
• OpenELM 1 1B Instruct — 1.1B, apple · $0.11/1M API est.
• PowerMoE 3b — 3.4B, ibm-research · $0.13/1M API est.
• Phi 4 mini instruct — 3.8B, microsoft · $0.22/1M API last-known
• Qwen2.5 1.5B — 1.5B, Qwen · $0.11/1M API est.
• h2ovl mississippi 800m — 800M, h2oai · $0.11/1M API est.
• h2ovl mississippi 2b — 2.2B, h2oai · $0.12/1M API est.
• Nomic Embed Text v1.5 — 137M, Nomic AI · $0.10/1M API est.
• Qwen2 0.5B — 500M, Qwen · $0.10/1M API est.
• SmolLM 1.7B Instruct quantized.w4a16 — 1.8B, nm-testing · $0.11/1M API est.
• Qwen2.5 1.5B quantized.w8a8 — 1.8B, RedHatAI · $0.11/1M API est.
• Qwen2.5 Math 1.5B — 1.5B, Qwen · $0.11/1M API est.
• gpt neo 2.7B — 2.7B, EleutherAI · $0.12/1M API est.
• Llama 3.2 3B — 3.2B, meta-llama · $0.13/1M API est.
• Qwen2 0.5B Instruct — 500M, Qwen · $0.10/1M API est.
• Phi tiny MoE instruct — 3.8B, microsoft · $0.13/1M API est.
• Qwen2.5 Coder 1.5B Instruct — 1.5B, Qwen · $0.11/1M API est.
• Qwen2.5 Coder 3B — 3.1B, Qwen · $0.12/1M API est.
• Kimi K3 DSpark — 2.2B, RadixArk · $0.12/1M API est.
• Qwen3 1.7B Base — 1.7B, Qwen · $0.11/1M API est.
• DeepSeek R1 Distill Qwen 1.5B — 1.8B, deepseek-ai · $0.11/1M API est.
• Phi 3 mini 4k instruct — 3.8B, microsoft · $0.13/1M API est.
• Qwen3 0.6B Base — 600M, Qwen · $0.10/1M API est.
• SmolLM3 3B — 3.1B, HuggingFaceTB · $0.12/1M API est.
• Qwen2.5 3B — 3.1B, Qwen · $0.12/1M API est.
• bloom 560m — 600M, bigscience · $0.10/1M API est.
• Qwen3Guard Gen 0.6B — 800M, Qwen · $0.11/1M API est.
• phi 2 — 2.8B, microsoft · $0.12/1M API est.
• gpt2 medium — 400M, openai-community · $0.10/1M API est.
• Llama 3.2 1B Instruct — 1.2B, unsloth · $0.11/1M API est.
• Zamba2 1.2B instruct — 1.2B, Zyphra · $0.11/1M API est.
• OLMo 2 0425 1B — 1.5B, allenai · $0.11/1M API est.
• SmolLM2 360M Instruct — 400M, HuggingFaceTB · $0.10/1M API est.
• gemma 2 2b it — 2.6B, google · $0.12/1M API est.
• LLaMmlein 1B prerelease — 1.1B, LSX-UniWue · $0.11/1M API est.
• t5gemma s s prefixlm — 300M, google · $0.10/1M API est.
• Ilama 3.2 1B — 1.2B, hmellor · $0.11/1M API est.
• SmolLM3 3B Base — 3.1B, HuggingFaceTB · $0.12/1M API est.
• granite 4.1 3b — 3.4B, ibm-granite · $0.13/1M API est.
• SmolLM2 1.7B — 1.7B, HuggingFaceTB · $0.11/1M API est.
• gemma 2b — 2.5B, google · $0.12/1M API est.
• Phi 3 mini 128k instruct — 3.8B, microsoft · $0.13/1M API est.
• ReaderLM v2 — 1.5B, jinaai · $0.11/1M API est.
• gemma 2 2b it — 2.6B, Efficient-Large-Model · $0.12/1M API est.
• Qwen2.5 Math 1.5B Instruct — 1.5B, Qwen · $0.11/1M API est.
• Qwen2.5 Coder 1.5B — 1.5B, Qwen · $0.11/1M API est.
• gemma 2 2b — 2.6B, google · $0.12/1M API est.
• bloomz 560m — 600M, bigscience · $0.10/1M API est.
• MiniCPM5 1B — 1.1B, openbmb · $0.11/1M API est.
• HRM Text 1B — 1.2B, sapientinc · $0.11/1M API est.
• PowerLM 3b — 3.5B, ibm-research · $0.13/1M API est.
• gpt2 xl — 1.6B, openai-community · $0.11/1M API est.
• gemma 1.1 2b it — 2.5B, google · $0.12/1M API est.
• SmolLM2 1.7B Instruct — 1.7B, HuggingFaceTB · $0.11/1M API est.
• LFM2.5 1.2B Instruct — 1.2B, LiquidAI · $0.11/1M API est.
• Midm 2.0 Mini Instruct — 2.3B, K-intelligence · $0.12/1M API est.
• Qwen3 4B Instruct 2507 NVFP4 — 2.8B, llmat · $0.12/1M API est.
• gemma 3 270m it — 300M, google · $0.10/1M API est.
• Qwen2.5 Coder 3B Instruct — 3.1B, Qwen · $0.12/1M API est.
• pythia 410m — 500M, EleutherAI · $0.10/1M API est.
• kanana nano 2.1b embedding — 2.1B, kakaocorp · $0.12/1M API est.
• Qwen2 1.5B — 1.5B, Qwen · $0.11/1M API est.
• LFM2 1.2B — 1.2B, LiquidAI · $0.11/1M API est.
• starcoder2 3b — 3B, bigcode · $0.12/1M API est.
• functiongemma 270m it — 300M, google · $0.10/1M API est.
• vlt5 base keywords — 300M, Voicelab · $0.10/1M API est.
• Qwen3.5 27B DFlash — 2.1B, z-lab · $0.12/1M API est.
• Falcon H1 0.5B Base — 500M, tiiuae · $0.10/1M API est.
• HyperCLOVAX SEED Vision Instruct 3B — 3.7B, naver-hyperclovax · $0.13/1M API est.
• Qwen2.5 Coder 0.5B Instruct — 500M, Qwen · $0.10/1M API est.
• pythia 1b — 1.1B, EleutherAI · $0.11/1M API est.
• Nemotron Labs Diffusion 3B Base — 3.8B, nvidia · $0.13/1M API est.
• Qwen1.5 0.5B — 600M, Qwen · $0.10/1M API est.
• Qwen1.5 1.8B — 1.8B, Qwen · $0.11/1M API est.
• EXAONE 3.5 2.4B Instruct — 2.4B, LGAI-EXAONE · $0.12/1M API est.
• MiniCPM4 0.5B — 400M, openbmb · $0.10/1M API est.
• pythia 1.4b — 1.5B, EleutherAI · $0.11/1M API est.
• gemma 3 1b pt — 1B, google · $0.11/1M API est.
• SmolLM 1.7B — 1.7B, HuggingFaceTB · $0.11/1M API est.
• LFM2.5 2.6B — 2.7B, LiquidAI · $0.12/1M API est.
• Qwen2.5 Coder 1.5B Instruct — 1.5B, unsloth · $0.11/1M API est.
• Qwen3.8 27B DFlash2 — 1.9B, incoai · $0.12/1M API est.
• Llama 3.2 3B Instruct — 3.2B, unsloth · $0.13/1M API est.
• qwen sft countdown defaultproj — 500M, asingh15 · $0.10/1M API est.
• Llama 3.2 1B — 1.2B, unsloth · $0.11/1M API est.
• stablelm 3b 4e1t — 2.8B, stabilityai · $0.12/1M API est.
• granite 3.0 1b a400m base — 1.4B, ibm-granite · $0.11/1M API est.
• Qwen3.6 35B A3B DFlash — 500M, z-lab · $0.10/1M API est.
• pythia 410m deduped — 500M, EleutherAI · $0.10/1M API est.
• Llama 3.2 3B Instruct pythonic — 3.2B, baseten · $0.13/1M API est.
• Qwen3.6 27B DFlash — 1.7B, z-lab · $0.11/1M API est.
• Qwen3 8B DFlash b16 — 1B, z-lab · $0.11/1M API est.
• granite 4.0 micro — 3.4B, ibm-granite · $0.13/1M API est.
• SmolLM2 360M — 400M, HuggingFaceTB · $0.10/1M API est.
• gemma 3 1b it — 1B, unsloth · $0.11/1M API est.
• Qwen3 8B speculator.eagle3 — 1B, RedHatAI · $0.11/1M API est.
• Qwen1.5 0.5B Chat — 600M, Qwen · $0.10/1M API est.
• LFM2.5 350M — 400M, LiquidAI · $0.10/1M API est.
• Qwen1.5 1.8B Chat — 1.8B, Qwen · $0.11/1M API est.
• Phi 4 mini reasoning — 3.8B, microsoft · $0.13/1M API est.
• tinyllama oneshot w8w8 test static shape change — 1.1B, nm-testing · $0.11/1M API est.
• kimi k2.6 eagle3 mla — 3B, lightseekorg · $0.12/1M API est.
• Ouro 1.4B — 1.4B, ByteDance · $0.11/1M API est.
• TwIL LM3 — 3.1B, webAI-Official · $0.12/1M API est.
• TinyLlama 1.1B intermediate step 1431k 3T — 1.1B, TinyLlama · $0.11/1M API est.
• gemma 2b it — 2.5B, google · $0.12/1M API est.
• deepseek coder 1.3b instruct — 1.3B, deepseek-ai · $0.11/1M API est.
• Qwen3.8 27B DFlash2 — 1.9B, z-lab · $0.12/1M API est.
• phi 1 5 — 1.4B, microsoft · $0.11/1M API est.
• Nemotron Labs Diffusion 3B — 3.8B, nvidia · $0.13/1M API est.
• gpt oss 20b speculator.eagle3 — 900M, RedHatAI · $0.11/1M API est.
• Qwen3.8 27B DSpark — 1.4B, RadixArk · $0.11/1M API est.
• LFM2.5 350M Base — 400M, LiquidAI · $0.10/1M API est.
• VibeThinker 3B — 3.1B, WeiboAI · $0.12/1M API est.
• OneReason 0.8B pretrain competition — 800M, OpenOneRec · $0.11/1M API est.
• NVIDIA Nemotron 3.5 Lightning 30B A3B NVFP4 DSpark — 800M, nvidia · $0.11/1M API est.
• gemma 4 26B A4B it DFlash — 400M, z-lab · $0.10/1M API est.
• LFM2.5 1.2B Base — 1.2B, LiquidAI · $0.11/1M API est.
• TIPO 500M — 500M, KBlueLeaf · $0.10/1M API est.
• LFM2.5 1.2B Thinking — 1.2B, LiquidAI · $0.11/1M API est.
• Qwen3.5 122B A10B DFlash — 800M, z-lab · $0.11/1M API est.
• Kimi K3 DSpark — 3.6B, Inferact · $0.13/1M API est.
• Nanbeige4.1 3B — 3.9B, Nanbeige · $0.13/1M API est.
• gemma 4 31B it DFlash — 1.5B, z-lab · $0.11/1M API est.
• Ouro 2.6B Thinking — 2.7B, ByteDance · $0.12/1M API est.
• llama 3.2 Korean Bllossom 3B — 3.2B, Bllossom · $0.13/1M API est.
• Qwen3.5 4B DFlash — 600M, z-lab · $0.10/1M API est.
• antares 1b — 1.8B, fdtn-ai · $0.11/1M API est.
• MiniCPM5 1B SFT — 1.1B, openbmb · $0.11/1M API est.
• tiny aya base — 3.3B, CohereLabs · $0.13/1M API est.
• LFM2.5 1.2B JP 202606 — 1.2B, LiquidAI · $0.11/1M API est.
• GLM 5.2 speculator.dspark — 3.8B, RedHatAI · $0.13/1M API est.
• bitnet b1.58 2B 4T — 800M, microsoft · $0.11/1M API est.
• LFM2 350M — 400M, LiquidAI · $0.10/1M API est.
• LFM2.5 2.6B Base — 2.7B, LiquidAI · $0.12/1M API est.
• granite 4.0 350m — 400M, ibm-granite · $0.10/1M API est.
• tiny aya global — 3.3B, CohereLabs · $0.13/1M API est.
• MiniMax M3 EAGLE3 — 3.3B, Inferact · $0.13/1M API est.
• Mythos nano — 3.1B, squ11z1 · $0.12/1M API est.
• Ouro 2.6B — 2.7B, ByteDance · $0.12/1M API est.
• PapersRAG 1.5B — 1.5B, metaresearch · $0.11/1M API est.

Our shortlist: Best small open LLMs →

Other sizes

Flagship (80B+) · Large (34–80B) · Medium (14–34B) · Small (4–14B) · All models

Compare → · LLM cost calculator →

Open the free LLM cost calculator → · Rate table dated 2026-08-24. $0 markup, your own accounts, we never resell compute. © 2026 Cynosure LLC.