Local-Inference on accordingly

Local-Inference on accordingly https://accordingly.dev/en/tags/local-inference/ Recent content in Local-Inference on accordingly Hugo en-us Sun, 27 Sep 2026 09:00:00 +0200 190 tokens per second: a 35-billion-parameter model on a laptop https://accordingly.dev/en/posts/190-token-al-secondo-qwen36-35b-a3b-rtx5080/ Sun, 27 Sep 2026 09:00:00 +0200 https://accordingly.dev/en/posts/190-token-al-secondo-qwen36-35b-a3b-rtx5080/ A 14 GB quantized Qwen3.6-35B-A3B generates 190 tokens/s on a 16 GB RTX 5080 Laptop. MTP, q8_0 KV cache and the full numbers from the test campaign.