Overview
Live data · synced 44 min ago. Compare 490 models from 63 companies, see what runs on your own machine, and let the advisor cut your token bill.
Active models
490
44 deprecated
Companies
63
model creators
Run locally
147
151 open-weights
Cheapest / 1M
$0.022
gpt-oss-120b (3:1 in/out)
Largest context
1.05M
GPT Astra Latest
Best value by tier
All models →- DeepSeek V4 Pro 0813$1.15
- Nova Pro 1.0$1.40
- Hermes 4 405B$1.50
- Doubao Seed 2.1 Pro 260915$1.73
- Doubao Seed 2.1 Pro 260628$1.73
- Qwen3.5-122B-A10B$0.72
- Nemotron 3 Super$0.82
- Gemini Exp 1206$0.85
- Moondream 3 Preview$0.85
- R1 0528$0.91
- gpt-oss-120b$0.022
- Mistral Nemo$0.029
- Ling 3.0 Flash$0.032
- Llama 3 8B Lunaris$0.042
- Nex-N2.5-Mini$0.044
Cheapest by blended price per 1M tokens (3 input : 1 output).
Recent price changes
All updates →- Kimi Latest~moonshotai · 44 min agoin $0.16 → $0.156%
- DeepSeek V4 Pro 0813DeepSeek · 44 min agoin $0.12 → $0.117%out $4.60 → $4.287%
- DeepSeek Pro Latest~deepseek · 44 min agoin $0.10 → $0.0967%
- DeepSeek V4.1 Flash 1017DeepSeek · 44 min agoin $0.30 → $0.1550%out $1.20 → $0.6050%
- gpt-oss-120bOpenAI · 44 min agoin $0.014 → $0.0137%out $0.052 → $0.0488%
Models
Prices are USD per 1M tokens from vendor price lists. Est. cost / token is blended at a 3:1 input:output mix.
| Compare | Model | Company | Parameters | Context | Input $/1M | Output $/1M | Est. cost / token | Features |
|---|---|---|---|---|---|---|---|---|
| Llama 4 Maverick v4fastlocal | Meta | 400B (17B active) | 524K | $0.35 | $1.00 | $0.51 /1M $0.000000512/token | VisionCachingJSON schemaTools | |
| Llama 4 Scout v4fastlocal | Meta | 109B (17B active) | 328K | $0.10 | $0.30 | $0.15 /1M $0.00000015/token | VisionJSON schemaTools | |
| Llama 3.3 70B Instruct v3.3fastlocal | Meta | 70.6B | 131K | $0.20 | $0.52 | $0.28 /1M $0.00000028/token | CachingJSON schemaTools | |
| Llama 3.2 1B Instruct v3.2fastlocal | Meta | 1.2B | 60K | $0.027 | $0.20 | $0.071 /1M $0.000000071/token | ||
| Llama 3.2 3B Instruct v3.2fastlocal | Meta | 3.2B | 80K | $0.051 | $0.34 | $0.12 /1M $0.000000122/token | JSON schema | |
| Llama 3.1 70B Instruct v3.1fastlocal | Meta | 70.6B | 131K | $0.72 | $0.72 | $0.72 /1M $0.00000072/token | ||
| Llama 3.1 8B Instruct v3.1fastlocal | Meta | 8B | 131K | $0.050 | $0.080 | $0.058 /1M $0.000000057/token | CachingJSON schemaTools | |
| Muse Wonder 1.0 v1frontier | Meta | Undisclosed | 1.05M | $4.00 | $20.00 | $8.00 /1M $0.000008/token | VisionPDFCachingReasoningJSON schema+2 | |
| Phi 4 v4fastlocal | Microsoft | 14.7B | 16K | $0.070 | $0.14 | $0.088 /1M $0.000000087/token | JSON schema | |
| MiniMax M3.1 Flash v3.1fastlocal | MiniMax | 220.5B | 1.05M | $0.15 | $0.60 | $0.26 /1M $0.000000262/token | VisionCachingReasoningJSON schemaTools | |
| MiniMax M3 v3fastlocal | MiniMax | 427B | 1M | $0.30 | $1.20 | $0.52 /1M $0.000000525/token | VisionCachingReasoningJSON schemaTools | |
| MiniMax M2.5 v2.5fastlocal | MiniMax | 228.7B | 1M | $0.30 | $1.20 | $0.52 /1M $0.000000525/token | CachingReasoningJSON schemaTools | |
| MiniMax M2.1 v2.1fastlocal | MiniMax | 228.7B | 1M | $0.30 | $1.20 | $0.52 /1M $0.000000525/token | CachingReasoningJSON schemaTools | |
| MiniMax M2 v2fastlocal | MiniMax | 228.7B | 200K | $0.30 | $1.20 | $0.52 /1M $0.000000525/token | CachingReasoningTools | |
| MiniMax M1 v1balanced | MiniMax | Undisclosed | 1M | $0.55 | $2.20 | $0.96 /1M $0.000000963/token | ReasoningTools | |
| MiniMax M2 1 Lightning v2.1fast | MiniMax | Undisclosed | 1M | $0.30 | $2.40 | $0.82 /1M $0.000000825/token | CachingReasoningTools | |
| MiniMax M2 5 Lightning v2.5fast | MiniMax | Undisclosed | 1M | $0.30 | $2.40 | $0.82 /1M $0.000000825/token | CachingReasoningTools | |
| Mistral Large 4 v4balanced | Mistral AI | Undisclosed | 1.05M | $0.70 | $2.10 | $1.05 /1M $0.00000105/token | VisionCachingReasoningJSON schemaTools | |
| Mistral Medium 3.5 v3.5balanced | Mistral AI | Undisclosed | 262K | $1.50 | $7.50 | $3.00 /1M $0.000003/token | VisionReasoningJSON schemaTools | |
| Mistral Small 4 fastlocal | Mistral AI | 119.4B | 262K | $0.15 | $0.60 | $0.26 /1M $0.000000262/token | VisionCachingReasoningJSON schemaTools | |
| Devstral 2 2512 fastlocal | Mistral AI | 125B | 262K | $0.40 | $2.00 | $0.80 /1M $0.0000008/token | CachingJSON schemaTools | |
| Ministral 3 14B 2512 fastlocal | Mistral AI | 13.9B | 262K | $0.20 | $0.20 | $0.20 /1M $0.0000002/token | VisionCachingJSON schemaTools | |
| Ministral 3 8B 2512 fastlocal | Mistral AI | 8.9B | 262K | $0.15 | $0.15 | $0.15 /1M $0.00000015/token | VisionCachingJSON schemaTools | |
| Ministral 3 3B 2512 fastlocal | Mistral AI | 3.8B | 131K | $0.10 | $0.10 | $0.10 /1M $0.0000001/token | VisionCachingJSON schemaTools | |
| Mistral Large 3 2512 fast | Mistral AI | Undisclosed | 262K | $0.50 | $1.50 | $0.75 /1M $0.00000075/token | VisionCachingJSON schemaTools |
Local LLMs
Open-weights models you can run on your own hardware: what to install, whether it fits your machine, whether it beats paying per token, and a live link to Ollama / LM Studio.
| Run it | ||||||
|---|---|---|---|---|---|---|
| Kimi K3.1 Pro Moonshot AI | 2.78T | 1797 GB | — | Hugging Face | $2.00 / $10.00 | 1.05M |
| Kimi K3 Moonshot AI | 2.78T | 1797 GB | — | Hugging Face | $3.00 / $15.00 | 1.05M |
| Qwen3.8 2.4T A95B Alibaba (Qwen) | 2.45T (95B active) | 1582 GB | — | Hugging Face | $2.00 / $6.00 | 1.05M |
| LongCat 2.0 Meituan | 1.78T | 1151 GB | — | Hugging Face | $0.30 / $1.20 | 1.05M |
| DeepSeek V4 Pro 0813 DeepSeek | 1.65T | 1070 GB | — | Hugging Face | $0.11 / $4.28 | 1.05M |
| DeepSeek V4 Pro 0423 DeepSeek | 1.6T | 1037 GB | — | Hugging Face | $1.32 / $3.96 | 1M |
| Kimi K2.6 Moonshot AI | 1.03T | 668 GB | — | Hugging Face | $0.95 / $4.00 | 262K |
| Kimi K2.7 Code Moonshot AI | 1.03T | 668 GB | — | Hugging Face | $0.95 / $4.00 | 262K |
| Kimi K2 0711 Moonshot AI | 1.03T | 668 GB | — | Hugging Face | $0.57 / $2.30 | 131K |
| Kimi K2 0905 Moonshot AI | 1.03T | 668 GB | — | Hugging Face | $0.60 / $2.50 | 262K |
Showing 1–10 of 147 local models
Memory is an estimate for Q4_K_M at 8K context. “Min. phone RAM” assumes a 4K context and that an app may use about half of the phone’s RAM; “Phone-ready” means it runs on a phone with 8 GB or less. “ollama run” tags come from a curated map (never guessed); verify with ollama show.
Advisor
Which model to use for what, and how to combine several to cut token cost. Task scores are heuristics: validate on your own data.
Which LLM should I use?
Describe your workload: tasks, volume, tokens per request, latency, privacy or hardware.
Data sync
The scheduler refreshes prices every 6 hours; every run is recorded.
Models
490 / 534
active / total
Newest data
44 min ago
Last run
success
44 min ago
| Started | Trigger | Status | New | Changed | Price changes | Alerts | Errors |
|---|---|---|---|---|---|---|---|
| 2026-11-05 06:00 | cron | success | 0 | 10 | 5 | 15 | |
| 2026-11-05 03:20 | startup | success | 534 | 0 | 0 | 0 |