SDKs › ExLlamaV3: which chips it runs on
ExLlamaV3: which chips it runs on
ExLlamaV3 and the chips it runs on, with evidence and the date each fact was last checked.
Snapshot 2026-10-15 · 394 configs on AI boxes and 423 on GPU cards · 123 models. Every number is generated from the tracker snapshot and linked to where it was measured; nothing here is typed by hand.
turboderp (turboderp-org/exllamav3) · runtime. 20 chip entries: runs 11, partial 9.
- NVIDIA GeForce RTX 3060 12GB: yes · evidence (checked 2026-10-07)
- NVIDIA GeForce RTX 3090: yes · evidence (checked 2026-10-07)
- NVIDIA GeForce RTX 4060 Ti (16GB): yes · evidence (checked 2026-10-07)
- NVIDIA GeForce RTX 4070 Ti SUPER: yes · evidence (checked 2026-10-07)
- NVIDIA GeForce RTX 4080: yes · evidence (checked 2026-10-07)
- NVIDIA GeForce RTX 4080 SUPER: yes · evidence (checked 2026-10-07)
- NVIDIA GeForce RTX 5060 Ti (16GB): yes · evidence (checked 2026-10-07)
- NVIDIA GeForce RTX 5070 Ti: yes · evidence (checked 2026-10-07)
- NVIDIA GeForce RTX 5080: yes · evidence (checked 2026-10-07)
- NVIDIA RTX 4090 (24GB): yes · evidence (checked 2026-10-07)
- NVIDIA RTX 5090 (32GB): yes · evidence (checked 2026-10-07)
- NVIDIA A100 80GB: partial · evidence (checked 2026-10-07)
- NVIDIA B200: partial · evidence (checked 2026-10-07)
- NVIDIA B300 (Blackwell Ultra): partial · evidence (checked 2026-10-07)
- NVIDIA H100 SXM: partial · evidence (checked 2026-10-07)
- NVIDIA H200: partial · evidence (checked 2026-10-07)
- NVIDIA L40S: partial · evidence (checked 2026-10-07)
- NVIDIA RTX 6000 Ada (48GB): partial · evidence (checked 2026-10-07)
- NVIDIA RTX A6000: partial · evidence (checked 2026-10-07)
- NVIDIA RTX PRO 6000 Blackwell (96GB): partial · evidence (checked 2026-10-07)
More: all configs · hardware · news · contributors · methodology · llms.txt · llms-full.txt · API (OpenAPI) · JSON snapshot. Built by Altronis, private on-prem AI, Singapore.