Model catalog

Pick a category, pick a model, and deploy in one click.

Chat

Conversation and reasoning

59 models

DeepSeek R1 Distill · Llama 8B
8B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 33–48 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 122–175 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8 · needs 8 GB · est. 111–158 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8 · needs 8 GB
  Unavailable
16 more
DeepSeek R1 Distill · Qwen 1.5B
1.5B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 4 GB · est. 146–208 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 4 GB · est. 541–767 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8, FP16 · needs 4 GB · est. 489–694 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8, FP16 · needs 4 GB
  Unavailable
16 more
DeepSeek R1 Distill · Qwen 14B
14B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8 · needs 11 GB · est. 18–26 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8 · needs 11 GB · est. 67–96 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M · needs 11 GB · est. 60–87 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M · needs 11 GB
  Unavailable
16 more
DeepSeek R1 Distill · Qwen 32B
32B
vLLM · BF16
No hardware in this list has enough memory.
View details
DeepSeek R1 Distill · Qwen 32B
32B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 8–12 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 30–44 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 28–41 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
DeepSeek R1 Distill · Qwen 7B
7B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 34–50 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 129–183 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8 · needs 8 GB · est. 116–166 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8 · needs 8 GB
  Unavailable
16 more
Gemma 3 12B
12B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8 · needs 11 GB · est. 22–32 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8 · needs 11 GB · est. 82–118 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M · needs 11 GB · est. 74–107 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8 · needs 11 GB
  Unavailable
16 more
Gemma 3 270M
270M
llama.cpp · Q8_0
- M4 Pro · 24 GB unifiedBest value
  Q8_0, Q4, FP16 · needs 4 GB · est. 561–796 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q8_0, Q4, FP16 · needs 4 GB · est. 2074–2939 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q8_0, Q4, FP16 · needs 4 GB · est. 1876–2659 tok/s
  Unavailable
- M1 · 16 GB unified
  Q8_0, Q4, FP16 · needs 4 GB
  Unavailable
16 more
Gemma 3 27B
27B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 9–15 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 36–52 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 33–49 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
Gemma 3 4B
4B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 65–94 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 242–345 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 219–312 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8, FP16 · needs 8 GB
  Unavailable
16 more
Gemma 3n E4B
llama.cpp · Q8_0
- M4 Pro · 24 GB unifiedBest value
  Q8_0 · needs 11 GB · est. 22–32 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q8_0 · needs 11 GB · est. 82–117 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q8_0 · needs 11 GB · est. 74–106 tok/s
  Unavailable
- M1 · 16 GB unified
  Q8_0 · needs 11 GB
  Unavailable
16 more
Gemma 4 12B
12B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8 · needs 15 GB · est. 22–33 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8 · needs 15 GB · est. 84–121 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8 · needs 15 GB
  Unavailable
- M1 Pro · 16 GB unified
  Q4_K_M, Q8 · needs 15 GB · est. 16–24 tok/s
  Unavailable
15 more
Gemma 4 26B A4B
26B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 9–14 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 35–51 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 33–47 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
Gemma 4 31B
31B
llama.cpp · Q4_K_M
- M4 Max · 48 GB unifiedFastest
  Q4_K_M, Q8 · needs 44 GB · est. 13–20 tok/s
  Unavailable
- M3 Max · 48 GB unified
  Q4_K_M, Q8 · needs 44 GB · est. 9–14 tok/s
  Unavailable
- M4 Pro · 48 GB unified
  Q4_K_M, Q8 · needs 44 GB · est. 8–13 tok/s
  Unavailable
- Instinct MI210 · 64 GB
  Q4_K_M, Q8 · needs 44 GB
  Unavailable
1 more
View details
GLM-4 32B 0414
32B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 8–12 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 30–44 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 28–41 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
gpt-oss-120b
120B
vLLM · MXFP4
No hardware in this list has enough memory.
View details
gpt-oss-120b MXFP4
120B
llama.cpp · MXFP4
No hardware in this list has enough memory.
View details
gpt-oss-20b
20B
llama.cpp · MXFP4
- RTX 4090 · 24 GBFastest
  MXFP4, Q8 · needs 15 GB · est. 49–71 tok/s
  Unavailable
- RTX 3090 · 24 GB
  MXFP4, Q8 · needs 15 GB · est. 46–66 tok/s
  Unavailable
View details
gpt-oss-20b
20B
vLLM · MXFP4
- RTX 4090 · 24 GBFastest
  MXFP4, Q8 · needs 22 GB · est. 43–63 tok/s
  Unavailable
- RTX 3090 · 24 GB
  MXFP4, Q8 · needs 22 GB · est. 40–58 tok/s
  Unavailable
View details
Llama 3.1 8B
8B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 33–48 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 122–175 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8 · needs 8 GB · est. 111–158 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8 · needs 8 GB
  Unavailable
16 more
Llama 3.2 1B
1B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 4 GB · est. 202–288 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 4 GB · est. 748–1061 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8, FP16 · needs 4 GB · est. 677–960 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8, FP16 · needs 4 GB
  Unavailable
16 more
Llama 3.2 3B
3B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 81–115 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 299–425 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 270–384 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8, FP16 · needs 8 GB
  Unavailable
16 more
Llama 3.3 70B
70B
vLLM · AWQ
No hardware in this list has enough memory.
View details
Llama 3.3 70B
70B
llama.cpp · Q4_K_M
- M4 Max · 48 GB unifiedFastest
  Q4_K_M · needs 44 GB · est. 5–9 tok/s
  Unavailable
- M3 Max · 48 GB unified
  Q4_K_M · needs 44 GB · est. 4–6 tok/s
  Unavailable
- M4 Pro · 48 GB unified
  Q4_K_M · needs 44 GB · est. 3–6 tok/s
  Unavailable
- Instinct MI210 · 64 GB
  Q4_K_M · needs 44 GB
  Unavailable
1 more
View details
Magistral Small 2509
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 11–17 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 42–60 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 39–56 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
Mistral 7B v0.3
7B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 37–54 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 138–196 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8 · needs 8 GB · est. 125–178 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8 · needs 8 GB
  Unavailable
16 more
Mistral Small 3.2 24B
24B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 11–17 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 42–60 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 39–56 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
OLMo 2 7B Instruct
7B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 36–52 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 135–192 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8 · needs 8 GB · est. 122–174 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8 · needs 8 GB
  Unavailable
16 more
Olmo 3 7B Instruct
7B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 11 GB · est. 36–52 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 11 GB · est. 135–192 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8 · needs 11 GB · est. 122–174 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8 · needs 11 GB
  Unavailable
16 more
Olmo 3.1 32B Instruct
32B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 8–12 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 31–44 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 28–41 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
Olmo 3.1 32B Think
32B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 8–12 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 31–44 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 28–41 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
OLMoE 1B 7B Instruct
1B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 38–56 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 143–204 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 129–184 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8, FP16 · needs 8 GB
  Unavailable
16 more
Phi-3.5 mini Instruct
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 8 GB · est. 68–97 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 8 GB · est. 252–359 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M · needs 8 GB · est. 228–324 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M · needs 8 GB
  Unavailable
16 more
Phi-4 14B
14B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8 · needs 11 GB · est. 18–26 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8 · needs 11 GB · est. 66–95 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M · needs 11 GB · est. 60–86 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M · needs 11 GB
  Unavailable
16 more
Phi-4 mini 3.8B
3.8B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 65–94 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 242–344 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 219–312 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8, FP16 · needs 8 GB
  Unavailable
16 more
Phi-4 reasoning 14B
14B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8 · needs 11 GB · est. 18–26 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8 · needs 11 GB · est. 66–95 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M · needs 11 GB · est. 60–86 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M · needs 11 GB
  Unavailable
16 more
Qwen2.5 14B Instruct
14B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8 · needs 11 GB · est. 18–26 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8 · needs 11 GB · est. 67–96 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M · needs 11 GB · est. 60–87 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M · needs 11 GB
  Unavailable
16 more
Qwen2.5 32B Instruct
32B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 8–12 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 30–44 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 28–41 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
Qwen2.5 7B Instruct
7B
Container
- RTX 3090 · 24 GB
  needs 22 GB
  Unavailable
- RTX 4090 · 24 GB
  needs 22 GB
  Unavailable
View details
Qwen2.5 7B Instruct
7B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 34–50 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 129–183 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8 · needs 8 GB · est. 116–166 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8 · needs 8 GB
  Unavailable
16 more
Qwen3 1.7B
1.7B
llama.cpp · Q8_0
- M4 Pro · 24 GB unifiedBest value
  Q8_0, Q4, FP16 · needs 8 GB · est. 89–127 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q8_0, Q4, FP16 · needs 8 GB · est. 329–468 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q8_0, Q4, FP16 · needs 8 GB · est. 298–423 tok/s
  Unavailable
- M1 · 16 GB unified
  Q8_0, Q4, FP16 · needs 8 GB
  Unavailable
16 more
Qwen3 14B
14B
vLLM · AWQ
- RTX 4090 · 24 GBFastest
  AWQ, Q8 · needs 15 GB · est. 60–86 tok/s
  Unavailable
- RTX 3090 · 24 GB
  AWQ, Q8 · needs 15 GB · est. 56–80 tok/s
  Unavailable
View details
Qwen3 14B
14B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8 · needs 11 GB · est. 18–26 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8 · needs 11 GB · est. 67–96 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M · needs 11 GB · est. 60–87 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M · needs 11 GB
  Unavailable
16 more
Qwen3 30B A3B Instruct 2507
30B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 8–13 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 32–47 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 30–43 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
Qwen3 30B A3B
30B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 8–13 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 32–47 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 30–43 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
Qwen3 32B
32B
vLLM · AWQ
No hardware in this list has enough memory.
View details
Qwen3 32B
32B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 8–12 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 30–44 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 28–41 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
Qwen3 4B Instruct 2507
4B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 65–93 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 242–344 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 219–311 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8, FP16 · needs 8 GB
  Unavailable
16 more
Qwen3 4B
4B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 65–93 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 242–344 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 219–311 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8, FP16 · needs 8 GB
  Unavailable
16 more
Qwen3 4B Thinking 2507
4B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 65–93 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 242–344 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 219–311 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8, FP16 · needs 8 GB
  Unavailable
16 more
Qwen3 8B
8B
Container
- RTX 3090 · 24 GB
  needs 22 GB
  Unavailable
- RTX 4090 · 24 GB
  needs 22 GB
  Unavailable
View details
Qwen3 8B
8B
vLLM · AWQ
- RTX 4090 · 24 GBFastest
  AWQ, Q8, FP16 · needs 11 GB · est. 98–141 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  AWQ, Q8 · needs 11 GB · est. 89–127 tok/s
  Unavailable
- RTX 3090 · 24 GB
  AWQ, Q8, FP16 · needs 11 GB · est. 91–131 tok/s
  Unavailable
View details
Qwen3 8B
8B
vLLM · BF16
- RTX 4090 · 24 GBFastest
  BF16, Q4, Q8 · needs 22 GB · est. 36–53 tok/s
  Unavailable
- RTX 3090 · 24 GB
  BF16, Q4, Q8 · needs 22 GB · est. 34–49 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  BF16, Q4, Q8 · needs 22 GB
  Unavailable
- Instinct MI210 · 64 GB
  BF16, Q4, Q8 · needs 22 GB
  Unavailable
View details
Qwen3 Next 80B A3B Instruct
80B
llama.cpp · Q4_K_M
No hardware in this list has enough memory.
View details
QwQ 32B
32B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M · needs 22 GB · est. 8–12 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M · needs 22 GB · est. 30–44 tok/s
  Unavailable
- RTX 3090 · 24 GB
  Q4_K_M · needs 22 GB · est. 28–41 tok/s
  Unavailable
- RX 7900 XTX · 24 GB
  Q4_K_M · needs 22 GB
  Unavailable
7 more
Seed-OSS 36B Instruct
36B
llama.cpp · Q4_K_M
- M4 Max · 48 GB unifiedFastest
  Q4_K_M, Q8 · needs 30 GB · est. 11–17 tok/s
  Unavailable
- M1 Max · 32 GB unified
  Q4_K_M · needs 30 GB · est. 11–16 tok/s
  Unavailable
- M2 Max · 32 GB unified
  Q4_K_M · needs 30 GB · est. 11–16 tok/s
  Unavailable
- M3 Max · 48 GB unified
  Q4_K_M, Q8 · needs 30 GB · est. 8–12 tok/s
  Unavailable
3 more
SmolLM2 1.7B Instruct
1.7B
llama.cpp · Q8_0
- M4 Pro · 24 GB unifiedBest value
  Q8_0, Q4, FP16 · needs 8 GB · est. 89–128 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q8_0, Q4, FP16 · needs 8 GB · est. 332–471 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q8_0, Q4, FP16 · needs 8 GB · est. 300–426 tok/s
  Unavailable
- M1 · 16 GB unified
  Q8_0, Q4, FP16 · needs 8 GB
  Unavailable
16 more
SmolLM3 3B
3B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 4 GB · est. 85–122 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 4 GB · est. 315–448 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8, FP16 · needs 4 GB · est. 285–405 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8, FP16 · needs 4 GB
  Unavailable
16 more
Yi 1.5 9B Chat
9B
llama.cpp · Q4_K_M
- M4 Pro · 24 GB unifiedBest value
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 30–44 tok/s
  Deployable now · $0.12/h
- RTX 4090 · 24 GBFastest
  Q4_K_M, Q8, FP16 · needs 8 GB · est. 113–161 tok/s
  Unavailable
- RTX 3080 Ti · 12 GB
  Q4_K_M, Q8 · needs 8 GB · est. 102–146 tok/s
  Unavailable
- M1 · 16 GB unified
  Q4_K_M, Q8 · needs 8 GB
  Unavailable
16 more

Not sure what runs on your machine? See what fits.

Model catalog

💬 Chat

Deploy any Hugging Face model

Chat