FreeLLMBest Free LLMs, Ranked by Real Votes
Every LLM you can use without paying, ranked by blind human votes on Arena: free APIs, free chats from the labs, and open models you can download. Updated every night; data of 25 Oct 2026.
The best free LLM right now
Arena chat ranking of 22 Oct 2026: 227 of 418 models are free to use.
Kimi K3 (Max)
#1 among free LLMs · Arena score 1487 · #18 of 418 in Chat, paid models included
Lab chat
GLM-5.3 (Max)
#2 among free LLMs · Arena score 1481 · #26 of 418 in Chat, paid models included
Lab chat
MiMo-V2.6-Pro
#3 among free LLMs · Arena score 1480 · #28 of 418 in Chat, paid models included
—
Free LLM APIs you can call today
16 models on OpenRouter cost nothing per token today. Try any of them on Try and compare, with your own free OpenRouter account.
Flint
No Arena chat votes yet · 262K tokens of context
Free API · try it Model page
Xing4.0 29B A4B
No Arena chat votes yet · 262K tokens of context
Free API · try it Model page
NVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4)
Arena chat score 1413 · #149 of 418 · 1.0M tokens of context
Free API · try it Model page
Dots Studio: Dots3-Note Preview
No Arena chat votes yet · 512K tokens of context
Free API · try it Model page
LFM2.5-2.6B
No Arena chat votes yet · 66K tokens of context
Free API · try it Model page
NVIDIA-Nemotron-3.5-Lightning-30B-A3B (NVFP4)
Arena chat score 1353 · #216 of 418 · 1.0M tokens of context
Free API · try it Model page
Inkling-Small
Arena chat score 1405 · #158 of 418 · 1.0M tokens of context
Free API · try it Model page
Poolside: Laguna S 2.1
No Arena chat votes yet · 262K tokens of context
Free API · try it Model page
Inkling
Arena chat score 1442 · #94 of 418 · 1.0M tokens of context
Free API · try it Model page
Poolside: Laguna XS 2.1
No Arena chat votes yet · 262K tokens of context
Free API · try it Model page
Cohere: North Mini Code
No Arena chat votes yet · 256K tokens of context
Free API · try it Model page
NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4)
Arena chat score 1426 · #122 of 418 · 1.0M tokens of context
Free API · try it Model page
NVIDIA: Nemotron 3 Nano Omni
No Arena chat votes yet · 256K tokens of context
Free API · try it Model page
Gemma 4 26B A4B
Arena chat score 1437 · #100 of 418 · 262K tokens of context
Free API · try it Model page
Gemma 4 31B
Arena chat score 1452 · #76 of 418 · 262K tokens of context
Free API · try it Model page
NVIDIA-Nemotron-3-Super-120B-A12B
Arena chat score 1360 · #208 of 418 · 262K tokens of context
Free API · try it Model page
Free LLM chats from the labs
Most labs that release open models also run a free chat website. Here is each lab’s best free chat model on Arena.
DeepSeek
Best free model: DeepSeek-V4.1-Flash-20260910 (Max) · #34 of 418 in Chat
chat.deepseek.com
Kimi (Moonshot)
Best free model: Kimi K3 (Max) · #18 of 418 in Chat
www.kimi.com
GLM (Z.ai)
Best free model: GLM-5.3 (Max) · #26 of 418 in Chat
chat.z.ai
Qwen (Alibaba)
Best free model: Qwen3.5-397B-A17B · #92 of 418 in Chat
chat.qwen.ai
MiniMax
Best free model: MiniMax-M3 · #97 of 418 in Chat
agent.minimax.io
Tencent
Best free model: Hy3 · #74 of 418 in Chat
yuanbao.tencent.com
Best free model: Gemma 4 31B · #76 of 418 in Chat
aistudio.google.com
NVIDIA
Best free model: NVIDIA-Nemotron-3-Ultra-550B-A55B (NVFP4) · #122 of 418 in Chat
build.nvidia.com
Open LLMs that run on one 24 GB GPU
Open chat models that fit fully on one RTX 4090 (24 GB) at 4-bit with 8K of context, by Arena score. Memory and speed are estimates. Simulate your own hardware.
Gemma 4 31B
Arena score 1452 · #76 of 418 in Chat, paid models included
22 GB at 4-bit · about 36 tok/s
Gemma 4 26B A4B
Arena score 1437 · #100 of 418 in Chat, paid models included
17 GB at 4-bit · about 152 tok/s
Qwen3.8-27B
Arena score 1437 · #102 of 418 in Chat, paid models included
18 GB at 4-bit · about 43 tok/s
Qwen3.5-27B
Arena score 1409 · #155 of 418 in Chat, paid models included
18 GB at 4-bit · about 43 tok/s
Qwen3.5-35B-A3B
Arena score 1394 · #172 of 418 in Chat, paid models included
23 GB at 4-bit · about 188 tok/s
GLM-4.7-Flash
Arena score 1365 · #203 of 418 in Chat, paid models included
20 GB at 4-bit · about 140 tok/s
NVIDIA-Nemotron-3.5-Lightning-30B-A3B (NVFP4)
Arena score 1353 · #216 of 418 in Chat, paid models included
20 GB at 4-bit · about 193 tok/s
Qwen3-32B
Arena score 1347 · #225 of 418 in Chat, paid models included
23 GB at 4-bit · about 34 tok/s
Gemma 3 12B IT
Arena score 1342 · #235 of 418 in Chat, paid models included
9 GB at 4-bit · about 87 tok/s
Granite 4.2 30B
Arena score 1340 · #237 of 418 in Chat, paid models included
21 GB at 4-bit · about 38 tok/s
QwQ-32B
Arena score 1336 · #242 of 418 in Chat, paid models included
23 GB at 4-bit · about 34 tok/s
Qwen3-30B-A3B
Arena score 1327 · #252 of 418 in Chat, paid models included
20 GB at 4-bit · about 163 tok/s
Molmo 2 8B
Arena score 1322 · #261 of 418 in Chat, paid models included
7 GB at 4-bit · about 108 tok/s
gpt-oss-20b
Arena score 1317 · #269 of 418 in Chat, paid models included
13 GB at 4-bit · about 171 tok/s
NVIDIA-Nemotron-3-Nano-30B-A3B (BF16)
Arena score 1314 · #273 of 418 in Chat, paid models included
20 GB at 4-bit · about 176 tok/s
Granite 4.1 8B
Arena score 1304 · #285 of 418 in Chat, paid models included
7 GB at 4-bit · about 105 tok/s
Gemma 3 4B IT
Arena score 1303 · #286 of 418 in Chat, paid models included
4 GB at 4-bit · about 212 tok/s
Granite 4.2 3B
Arena score 1290 · #292 of 418 in Chat, paid models included
4 GB at 4-bit · about 213 tok/s
Granite 4.2 8B
Arena score 1290 · #294 of 418 in Chat, paid models included
7 GB at 4-bit · about 105 tok/s
Best free LLMs for coding
Arena’s web development ranking of 23 Oct 2026: models build working pages from a prompt. 60 of 150 are free.
DeepSeek-V4.1-Flash-20261017 (Max)
#1 among free models · Arena score 1665 · #14 of 150 in Coding, paid models included
Lab chat
Kimi K3 (Max)
#2 among free models · Arena score 1655 · #17 of 150 in Coding, paid models included
Lab chat
MiMo-V2.6-Flash
#3 among free models · Arena score 1635 · #20 of 150 in Coding, paid models included
—
Hy4 preview
#4 among free models · Arena score 1631 · #21 of 150 in Coding, paid models included
Lab chat
Qwen3.8-Flash-Next
#5 among free models · Arena score 1629 · #22 of 150 in Coding, paid models included
Lab chat
Frequently asked questions
What is the best free LLM?
By blind Arena votes, the best free LLM on 22 Oct 2026 is Kimi K3 (Max), with a score of 1487: #18 of 418 chat models, paid ones included. For coding, the best free model is DeepSeek-V4.1-Flash-20261017 (Max) (#14 of 150).
Is there a free LLM API?
Yes. Today 16 models on OpenRouter cost nothing per token; the best ranked on Arena are Gemma 4 31B (#76), Inkling (#94), Gemma 4 26B A4B (#100). You can try them side by side on our Try and compare page, with your own free OpenRouter account.
Can I run a free LLM on my own PC?
Yes, if it fits in your GPU memory. On one RTX 4090 (24 GB), 19 open chat models fit fully on the card at 4-bit. The best by Arena votes is Gemma 4 31B (#76), with about 22 GB and 36 tokens per second. The Run at home page estimates memory and speed for your own card.
Share this page:XLinkedInRedditWhatsApp
Spotted a wrong number or a broken link? Report a mistake