Tesla P40 24G — what LLMs can it run?
86 of 112 indexed models fit comfortably in 24GB at 4K context, each at the highest-quality quant that still leaves headroom.
The short answer for Tesla P40 24G
24 GB, 346 GB/s GDDR5. 86 of 112 models in this index fit comfortably at 4K context; 89 load at all.
- Biggest that fits
- Seed-OSS 36B Instruct — 19.9 GB at AWQ INT4 — 4.1 GB spare at 4K, so longer context comes out of a thin margin.
- Room to grow
- ERNIE 4.5 21B-A3B — 14.1 GB at Q4_K_M — under 60% of the card, which leaves 10.0 GB for a long context window or a second process.
- Speed ceiling
- Llama 3.1 8B Instruct at Q4_K_M reads 4.6 GB of weights per token, so 346 GB/s puts a hard ceiling near 75 tok/s. That is arithmetic on two published numbers, not a benchmark — real throughput lands below it. No run on this card has been measured here, so there is nothing to compare the ceiling against.
- The wall you will hit
- Bandwidth, not capacity. The models on this page fit, but at 346 GB/s this card reads the whole weight set once per generated token — a 1,008 GB/s RTX 4090 holds exactly the same 24 GB and moves those bytes 2.9× faster. Expect the same quant to generate proportionally slower here.
32B · 28
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| Seed-OSS 36B Instruct36B | AWQ INT4 | 19.91 GB | +4.1 GB | ≤20Ceiling |
| Command R 35B35B | GPTQ INT4 | 18.97 GB | +5 GB | ≤21Ceiling |
| Yi 1.5 34B Chat34B | AWQ INT4 | 19 GB | +5 GB | ≤21Ceiling |
| Qwen3 32B Instruct32B | AWQ INT4 | 18.22 GB | +5.8 GB | ≤22Ceiling |
| Qwen2.5 32B Instruct32B | EXL2 3.5bpw | 15.96 GB | +8 GB | ≤26Ceiling |
| Qwen2.5-Coder 32B Instruct32B | AWQ INT4 | 18.08 GB | +5.9 GB | ≤22Ceiling |
| DeepSeek-R1-Distill-Qwen-32B32B | EXL2 3.5bpw | 15.96 GB | +8 GB | ≤26Ceiling |
| Nemotron 3.5 Lightning 30B-A3B30B-A3B | Q4_0 | 19.79 GB | +4.2 GB | ≤203Ceiling |
| Gemma 4 31B IT31B | Q4_K_M | 21.08 GB | +2.9 GB | 11–15Derived |
| Qwen3 30B-A3B Instruct30B-A3B | Q4_K_M | 19.73 GB | +4.3 GB | ≤200Ceiling |
| Qwen3-Coder 30B-A3B Instruct30B-A3B | Q4_K_M | 19.73 GB | +4.3 GB | ≤200Ceiling |
| Qwen3-VL 30B-A3B Instruct30B-A3B | Q4_K_M | 19.73 GB | +4.3 GB | ≤200Ceiling |
| North Mini Code 1.0 30B-A3B30B-A3B | Q4_K_M | 20.01 GB | +4 GB | ≤197Ceiling |
| GLM-4.7-Flash30B-A3B | Q4_K_M | 19.19 GB | +4.8 GB | ≤200Ceiling |
| Granite 4.2 30B30B | Q4_K_M | 19.61 GB | +4.4 GB | 12–15Derived |
| Muse Glimmer 30B30B | Q4_K_M | 17.66 GB | +6.3 GB | 12–16Derived |
| Qwen3.8 27B27B | Q4_K_M | 17.47 GB | +6.5 GB | 13–17Derived |
| Qwen3.6 27B27B | Q4_K_M | 19.07 GB | +4.9 GB | 12–15Derived |
| Gemma 3 27B IT27B | Q4_K_M | 18.37 GB | +5.6 GB | 12–17Derived |
| Gemma 2 27B Instruct27B | Q4_K_M | 18.81 GB | +5.2 GB | 13–17Derived |
| Gemma 4 26B-A4B IT26B-A4B | Q4_K_M | 16.39 GB | +7.6 GB | ≤150Ceiling |
| Mistral Small 24B Instruct24B | EXL2 4.65bpw | 15.26 GB | +8.7 GB | ≤26Ceiling |
| LFM2.5 24B-A2B24B-A2B | Q4_K_M | 15.27 GB | +8.7 GB | ≤245Ceiling |
| Devstral Small 1.1 24B24B | Q6_K | 20.91 GB | +3.1 GB | 11–14Derived |
| Magistral Small 1.2 24B24B | Q6_K | 20.91 GB | +3.1 GB | 11–14Derived |
| Codestral 22B22B | Q4_K_M | 15.03 GB | +9 GB | 15–20Derived |
| ERNIE 4.5 21B-A3B21B-A3B | Q4_K_M | 14.05 GB | +10 GB | ≤200Ceiling |
| GPT-OSS 20B21B MoE | MXFP4 | 12.76 GB | +11.2 GB | ≤175Ceiling |
14B · 19
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| InternLM2 20B Chat20B | Q5_K_M | 15.59 GB | +8.4 GB | 15–19Derived |
| DeepSeek-Coder-V2-Lite Instruct16B-A2.4B | Q8_0 | 17.56 GB | +6.4 GB | ≤143Ceiling |
| DeepSeek-V2-Lite Chat16B-A2.4B | Q4_K_M | 10.08 GB | +13.9 GB | ≤250Ceiling |
| StarCoder2 15B15B | Q4_K_M | 10.19 GB | +13.8 GB | 23–30Derived |
| Qwen3 14B Instruct14B | Q5_K_M | 11.65 GB | +12.4 GB | 20–26Derived |
| Qwen2.5 14B Instruct14B | Q5_K_M | 11.73 GB | +12.3 GB | 20–26Derived |
| DeepSeek-R1-Distill-Qwen-14B14B | EXL2 4.65bpw | 9.75 GB | +14.3 GB | ≤43Ceiling |
| Phi-4 14B14B | Q5_K_M | 11.77 GB | +12.2 GB | 20–26Derived |
| Phi-3 Medium 14B Instruct14B | Q6_K | 12.86 GB | +11.1 GB | 18–24Derived |
| Mistral Nemo 12B Instruct12B | Q6_K | 11.14 GB | +12.9 GB | 21–27Derived |
| Gemma 3 12B IT12B | Q5_K_M | 9.84 GB | +14.2 GB | 24–32Derived |
| Mellum2.1 12B-A2.5B Thinking12B-A2.5B | Q8_0 | 13.6 GB | +10.4 GB | ≤137Ceiling |
| Stable LM 2 12B Chat12B | Q4_K_M | 8.35 GB | +15.7 GB | 28–37Derived |
| Gemma 4 12B IT12B | Q8_0 | 13.82 GB | +10.2 GB | 16–22Derived |
| Llama 3.2 11B Vision Instruct11B | Q8_0 | 12.9 GB | +11.1 GB | 18–23Derived |
| Solar 10.7B Instruct11B | Q4_K_M | 7.6 GB | +16.4 GB | 32–42Derived |
| Falcon 3 10B Instruct10B | Q4_K_M | 7.28 GB | +16.7 GB | 33–44Derived |
| Gemma 2 9B Instruct9B | Q8_0 | 11.7 GB | +12.3 GB | 21–28Derived |
| Qwen3.5 9B9B | Q8_0 | 10.37 GB | +13.6 GB | 21–28Derived |
7B · 25
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| GLM-4-9B-Chat9B | Q8_0 | 10.16 GB | +13.8 GB | 22–29Derived |
| Qwen3-VL 8B Instruct8B | Q8_0 | 10.39 GB | +13.6 GB | 22–29Derived |
| Granite 4.2 8B8B | Q8_0 | 10.45 GB | +13.6 GB | 22–29Derived |
| Ministral 3 8B Instruct8B | Q8_0 | 10.02 GB | +14 GB | 23–30Derived |
| Qwen2-VL 7B Instruct7B | Q4_K_M | 5.49 GB | +18.5 GB | 41–55Derived |
| Granite 3.1 8B Instruct8B | Q4_K_M | 5.88 GB | +18.1 GB | 42–55Derived |
| Qwen3 8B Instruct8B | Q6_K | 7.64 GB | +16.4 GB | 31–41Derived |
| Llama 3.1 8B Instruct8B | Q8_0 | 9.47 GB | +14.5 GB | 24–32Derived |
| Nous Hermes 3 Llama 3.1 8B8B | EXL2 4.65bpw | 5.43 GB | +18.6 GB | ≤78Ceiling |
| Aya 23 8B8B | Q4_K_M | 5.64 GB | +18.4 GB | 43–56Derived |
| OpenChat 3.6 8B8B | EXL2 4.65bpw | 5.43 GB | +18.6 GB | ≤78Ceiling |
| DeepSeek-R1-Distill-Llama-8B8B | Q5_K_M | 6.51 GB | +17.5 GB | 36–48Derived |
| InternLM2 7B Chat7B | Q4_K_M | 5.45 GB | +18.6 GB | 44–58Derived |
| Qwen2.5 7B Instruct7B | Q6_K | 6.77 GB | +17.2 GB | 33–44Derived |
| Qwen2.5-Coder 7B Instruct7B | EXL2 4.65bpw | 4.87 GB | +19.1 GB | ≤82Ceiling |
| DeepSeek-R1-Distill-Qwen-7B7B | EXL2 4.65bpw | 4.87 GB | +19.1 GB | ≤82Ceiling |
| Gemma 4 E4B ITE4B | Q8_0 | 8.46 GB | +15.5 GB | 26–34Derived |
| OLMo 2 7B Instruct7B | Q8_0 | 10.3 GB | +13.7 GB | 27–35Derived |
| OLMo 3 7B Instruct7B | Q8_0 | 10.3 GB | +13.7 GB | 27–35Derived |
| WizardLM-2 7B7B | Q4_K_M | 5.14 GB | +18.9 GB | 47–62Derived |
| Mistral 7B Instruct v0.37B | Q6_K | 6.75 GB | +17.3 GB | 35–46Derived |
| Zephyr 7B Beta7B | Q6_K | 6.75 GB | +17.3 GB | 35–46Derived |
| Gemma 4 E2B ITE2B | Q8_0 | 5.2 GB | +18.8 GB | ≤74Ceiling |
| Qwen3.5 4B4B | Q8_0 | 4.97 GB | +19 GB | ≤79Ceiling |
| Gemma 3 4B IT4B | Q8_0 | 5.05 GB | +19 GB | ≤80Ceiling |
≤3B · 14
| Model | Quant | Est. VRAM | Headroom | tok/s |
|---|---|---|---|---|
| Qwen3 4B Instruct4B | Q6_K | 4.05 GB | +20 GB | ≤111Ceiling |
| Phi-4 Mini Instruct3.8B | Q8_0 | 4.81 GB | +19.2 GB | ≤89Ceiling |
| Phi-3.5 Mini Instruct3.8B | Q8_0 | 5.89 GB | +18.1 GB | ≤90Ceiling |
| Granite 4.2 3B3B | Q8_0 | 4.41 GB | +19.6 GB | ≤94Ceiling |
| Llama 3.2 3B Instruct3B | Q8_0 | 4.05 GB | +20 GB | ≤107Ceiling |
| Qwen2.5 3B Instruct3B | Q8_0 | 3.59 GB | +20.4 GB | ≤111Ceiling |
| SmolLM3 3B3B | Q8_0 | 3.73 GB | +20.3 GB | ≤111Ceiling |
| Gemma 2 2B Instruct2B | Q8_0 | 3.34 GB | +20.7 GB | ≤132Ceiling |
| Qwen3.5 2B2B | Q8_0 | 2.23 GB | +21.8 GB | ≤175Ceiling |
| Qwen3 1.7B Instruct1.7B | Q8_0 | 2.39 GB | +21.6 GB | ≤199Ceiling |
| Qwen2.5 1.5B Instruct1.5B | Q8_0 | 1.83 GB | +22.2 GB | ≤223Ceiling |
| Llama 3.2 1B Instruct1B | Q8_0 | 1.51 GB | +22.5 GB | ≤277Ceiling |
| Qwen3.5 0.8B0.8B | Q8_0 | 0.94 GB | +23.1 GB | ≤427Ceiling |
| Qwen2.5 0.5B Instruct0.5B | Q8_0 | 0.6 GB | +23.4 GB | ≤706Ceiling |
How this list is built
Each row is the lowest-perplexity-loss quant of that model whose estimated total — weights plus KV cache at 4K context plus activation buffer — uses at most 88% of the card. That is the calculator's "green" threshold, so every row here has real headroom rather than only just fitting. Raise the context length and the list shortens; the calculator lets you check any combination directly. The speed column is this card: a measured run, a range scaled from this card's own Llama 2 7B benchmark, or a bandwidth ceiling. A figure that only exists for the RTX 4090 is labelled as that card.
A further 3 models load but with no headroom to spare (up to 105% of VRAM) — the Quant Hub’s GPU chips count those too, which is why its number is higher.
Measured on this card
No benchmark runs in this index were recorded on a Tesla P40 24G. Every figure on this page is calculated from the model architecture and the quant level — treat them as estimates, not measurements.
Setup guides for this card
Step-by-step guides written for the Tesla P40 24G's runtime or memory size, with the commands and the checks that prove the model is running on the card.
Stepping up
A A100 40G (40GB) fits 7 more of the indexed models than this card. A100 40G →
Common questions
What is the best local LLM for a Tesla P40 24G?
For everyday use, ERNIE 4.5 21B-A3B at Q4_K_M — about 14.1 GB of the card's 24 GB at 4K context, leaving 10.0 GB for a longer window. If you want the largest thing that will load, that is Seed-OSS 36B Instruct at AWQ INT4 (19.9 GB, 4.1 GB spare). "Best" here means best fit for the memory budget — this index does not run task benchmarks, so it cannot tell you which model is smarter.
Can a Tesla P40 24G run Mixtral 8x7B Instruct?
Only just. Its smallest build here, AWQ INT4, needs about 24.9 GB against 24.0 GB usable — that loads with nothing else running, with no margin for a longer context window. It is not on the list above, which requires a model to stay inside 88% of usable capacity.
How many tokens per second does a Tesla P40 24G do on an 8B model at Q4?
This index has no measured run on this card, so it does not publish a figure. What can be stated from specifications: generating a token requires reading every weight once, Llama 3.1 8B Instruct at Q4_K_M is 4.6 GB of weights, and this card moves 346 GB/s — a ceiling near 75 tok/s. Batch size, context length, the runtime and how much of the model sits in cache all take you below it.
Tesla P40 24G or RTX 4090 for local LLMs?
They hold the same models: 24 GB against 24 GB fits 86 and 86 of 112 respectively at 4K. The difference is throughput — 346 GB/s against 1,008 GB/s, a 2.9× gap in how fast the weights can be read, which is what token generation is bound by. The RTX 4090 generates faster on any model both can hold.
86 of 112 indexed models fit comfortably in 24GB at 4K context, each at the highest-quality quant that still leaves headroom.
This page's figures change when the model or the runtime does.Last updated 2026-10-20 RSS → /feed.xml