llamAmpere

llama.cpp fork for significantly improved performance on Ampere generation Nvidia cards (especially RTX 3090 / 3090 Ti): example: 110+ tok/s over a 100K-token generation at temperature 1 for qwen3.8, supporting full 262K ctx. Supports low memory (12gb) configs for 200k+ ctx. 120+ TPS, 3300+ prefill for QFN 3.8.

JakeATX/llamAmpere

View on GitHub ↗

llama.cpp fork for significantly improved performance on Ampere generation Nvidia cards (especially RTX 3090 / 3090 Ti): example: 110+ tok/s over a 100K-token generation at temperature 1 for qwen3.8, supporting full 262K ctx. Supports low memory (12gb) configs for 200k+ ctx. 120+ TPS, 3300+ prefill for QFN 3.8.

llamAmpere grew +119.9% in the last 30 days, ranks 258th of 539 in Local Runtime / Inference Engine, actively maintained, last commit today.

Star trajectory

65 days tracked
2026-09-17409 stars today2026-11-22
Momentum

+119.9% in 30d, above the Local Runtime / Inference Engine median of +1.9%.

Maintenance

Last commit today, fresher than the median of 6d.

How it compares in Local Runtime / Inference Engine

llamAmpere
409
stars
+119.9% 30d
commit today
Local Runtime / Inference Engine median
360
stars
+1.9% 30d
commit ~6d ago
Leader · ollama
187.8k
stars
+1.9% 30d
commit today

Ranked 258th of 539 by stars.

Ranked by shared tags, then keywords, then stars
  1. Serve large Qwen models fast on the GPUs you actually own. Qwen3.8-27B on a single 24 GB card with vLLM: 127…

    +592 (+27%) in 30d · Committed 12d ago · Python · Apache-2.0

    Shares Inference Optimization, Local Runtime / Inference Engine

  2. Run large AI models on your own PC: Qwen3.8-Flash-Next on Strata's engine, and GLM 5.3 Flash, DeepSeek V4…

    +962 (+243%) in 30d · Committed 3d ago · C++ · MIT

    Shares Local Runtime / Inference Engine, Inference Optimization

  3. RDNA-native LLM inference engine in Rust.

    +95 (+13%) in 30d · Committed today · Rust · Other license

    Shares Local Runtime / Inference Engine, Inference Optimization

  4. Pure Rust + CUDA LLM inference engine — no PyTorch, OpenAI-compatible, serves Qwen3 to Kimi-K2

    +25 (+3.5%) in 30d · Committed yesterday · Rust · Apache-2.0

    Shares Inference Optimization, Local Runtime / Inference Engine

  5. 37 tok/s from Qwen3-Next-80B (48 GB) on a 32 GB laptop, 24 without drafts vs 2.8 for llama.cpp; reading 62 vs…

    +280 (+61%) in 30d · Committed yesterday · C · GPL-3.0

    Shares Local Runtime / Inference Engine, Inference Optimization

  6. Xolotl: run DeepSeek-V4.1-Flash, all 552B, as a local coding agent on one gaming PC (RTX 4090 + 110 GB RAM)…

    +264 (+90%) in 30d · Committed today · C++ · MIT

    Shares Local Runtime / Inference Engine, Inference Optimization

  7. Alzeo is an open-source installer that turns an NVIDIA DGX Spark, or any GB10 machine (ASUS, Dell, HP, Lenovo…

    +36 (+15%) in 30d · Committed 2d ago · Python · MIT

    Shares Local Runtime / Inference Engine, Inference Optimization

  8. Optimized SGLang runtime for Qwen3.8-27B FP8 with DFlash2 and Qwen3.8 Flash-Next NVFP4 with FR-Spec on one…

    +31 (+18%) in 30d · Committed today · Python · Apache-2.0

    Shares Local Runtime / Inference Engine, Inference Optimization

  9. run a 27b local ai on your gaming gpu, rtx 3060 and up, radeon rx 6600 and up in beta, 8gb or more, linux and…

    +31 (+19%) in 30d · Committed 35d ago · Python

    Shares Local Runtime / Inference Engine, Inference Optimization

  10. Qwen3.8-Flash-Next on 2× RTX 3090: with 128 GB RAM up to 4,191 tok/s prefill · 111.5 tok/s decode (131K…

    +22 (+18%) in 30d · Committed 37d ago · Python · Apache-2.0

    Shares Local Runtime / Inference Engine, Inference Optimization

Language
C++
License
MIT License
Contributors
2,277
Forks
51
Watchers
4
Open issues
8
Releases
31
Created
2026-09-07
Stars this week
+29

You've seen where llamAmpere stands today.

Get the moves before everyone else: the 5 Local Runtime / Inference Engine repos climbing fastest, every Tuesday.

Free weekly digest. Double opt-in, unsubscribe anytime.

Maintain llamAmpere?

Claim this page to add your own call-to-action.

Claim this page