memra-reference
Rust + CUDA LLM inference engine for Blackwell (Tuned specifically on RTX PRO 6000, RTX 5090, B200): OpenAI-compatible (+converse and ant) serving, per-model X hardware exactness gates. NVFP4/mixed (fp8 hybrid, 4o6, etc - correctness, performance, hardware specific adapted) main quant support.
View on Github
Activity
- Latest release
- 3d ago
- Total releases
- 40
- Cadence
- ~daily
- Last 12 months
- 40
Reach
- Downloads
- 850
- Stars
- 1
Details
- License
- unknown
- First release
- Aug 23, 2026
Releases
| Version | Released | |
|---|---|---|
0.138.0
minor
| ||
0.137.0
minor
| ||
0.136.0
minor
| ||
0.135.0
minor
| ||
0.134.0
minor
| ||
0.133.0
minor
| ||
0.132.0
minor
| ||
0.131.0
minor
| ||
0.130.0
minor
| ||
0.129.0
minor
| ||
0.128.0
minor
| ||
0.127.0
minor
| ||
0.126.1
patch
| ||
0.126.0
minor
| ||
0.125.0
minor
| ||
0.124.1
patch
| ||
0.124.0
minor
| ||
0.123.0
minor
| ||
0.122.0
minor
| ||
0.121.0
minor
| ||
0.120.0
minor
| ||
0.119.0
minor
| ||
0.118.0
minor
| ||
0.117.0
minor
| ||
0.116.1
patch
| ||
0.116.0
minor
| ||
0.115.0
minor
| ||
0.113.2
patch
| ||
0.113.1
patch
| ||
0.113.0
minor
| ||
0.112.1
patch
| ||
0.112.0
minor
| ||
0.111.0
minor
| ||
0.110.0
minor
| ||
0.109.2
patch
| ||
0.109.1
patch
| ||
0.109.0
minor
| ||
0.108.1
patch
| ||
0.108.0
minor
| ||
0.107.0
initial
|