Find the model. Know what runs.
We track 562 open-weight AI models from 38 labs: what shipped, the memory each one needs, the 53 GPUs and Macs that can hold it, and the cheapest hosted API in USD and INR.
The API figure is the most downloaded chat model with a hosted API, Qwen3-14B, at its cheapest offer on a 3:1 mix of input and output tokens.
Latest
Guides, China Desk, licenses, run guides and lab historiesGet the weekly list
Get the weekly list of new open models and price changes. No spam. Unsubscribe any time.
New this week
3 releases in the 7 days to 7 Nov 2026, the newest release in our dataChina Desk
Chinese labs shipped 38% of open releases in the last 12 months (78 of 205)| Month | IBM Granite | NVIDIA | Lightricks | Swiss AI Initiative | Z.ai (Zhipu) | Black Forest Labs | DeepSeek | Other labs | Total | |
|---|---|---|---|---|---|---|---|---|---|---|
| Dec 2025 | 0 | 2 | 0 | 0 | 2 | 0 | 1 | 0 | 13 | 18 |
| Jan 2026 | 0 | 0 | 6 | 0 | 1 | 5 | 1 | 0 | 9 | 22 |
| Feb 2026 | 1 | 0 | 0 | 1 | 1 | 0 | 0 | 0 | 12 | 15 |
| Mar 2026 | 1 | 3 | 3 | 1 | 0 | 0 | 0 | 5 | 6 | 19 |
| Apr 2026 | 5 | 0 | 0 | 1 | 1 | 1 | 2 | 1 | 11 | 22 |
| May 2026 | 0 | 1 | 0 | 3 | 0 | 0 | 0 | 2 | 10 | 16 |
| Jun 2026 | 0 | 3 | 0 | 0 | 2 | 0 | 1 | 1 | 12 | 19 |
| Jul 2026 | 0 | 2 | 2 | 2 | 0 | 0 | 1 | 0 | 6 | 13 |
| Aug 2026 | 5 | 2 | 0 | 1 | 4 | 0 | 2 | 0 | 8 | 22 |
| Sep 2026 | 0 | 0 | 1 | 2 | 0 | 0 | 1 | 1 | 10 | 15 |
| Oct 2026 | 3 | 2 | 0 | 0 | 0 | 4 | 1 | 0 | 11 | 21 |
| Nov 2026 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 3 | 3 |
- Releases, last 12 months
- 205
- Busiest month
- 22
- Aug 2026
- Most releases
- 15
- IBM Granite
Small models worth running
Most downloaded under 10B parameters, with fit at 4-bit
Qwen3-0.6B8 GB: Fits16 GB: Fits
Florence-2-base8 GB: Fits16 GB: Fits
Qwen3-Embedding-0.6B8 GB: Fits16 GB: Fits
Qwen3-VL-8B-Instruct8 GB: Fits16 GB: Fits
granite-embedding-small-english-r28 GB: Fits16 GB: Fits
Qwen2.5-7B-Instruct8 GB: Fits16 GB: Fits
Qwen3.5-9B8 GB: Fits16 GB: Fits
Qwen3-8B8 GB: Fits16 GB: Fits
What can my hardware run?
All 53 devicesMacBook Pro M5 Max (48GB)
48 GB unified36GB usable391 models fit at 4-bitMac Studio M5 Ultra (256GB)
256 GB unified192GB usable432 models fit at 4-bitGeForce RTX 5090
32 GB VRAM32GB usable391 models fit at 4-bitRTX PRO 6000 Blackwell
96 GB VRAM96GB usable417 models fit at 4-bitDGX Spark
128 GB unified108.8GB usable418 models fit at 4-bitMemory fit is an estimate, not a runtime test. Macs count 66% of memory as usable up to 36 GB and 75% above; AI boxes count 85% because the memory is shared with the system.
Pricing snapshot
Lowest blended price, chat models of 100B+ parameters
DeepSeek-V4-Flash$0.0314₹2.96 / 1M
gpt-oss-120b$0.045₹4.24 / 1M
GLM-5.3-Flash$0.0495₹4.67 / 1M
DeepSeek-V4.1-Flash-1017$0.0815₹7.69 / 1M
DeepSeek-V4-Flash-0731$0.0875₹8.25 / 1M
Blended is 3 parts input to 1 part output, USD per million tokens, with INR at ₹94.29 per $1.