NVIDIA · Pascal · 24 GB VRAM tier
Best AI Models for NVIDIA Tesla P40 (24 GB)
Runs 14B and 35B MoE models well; 27–31B fits but is slow (13 tok/s).
24 GB GDDR5346 GB/s12 TFLOPS FP16250 WPascal 2016
Pascal: no Tensor Cores and 1:64-rate FP16 — fine for GGUF/INT8 inference, poor for FP16/exllama. Passive server card; needs added cooling in a desktop.
| # | Model | Speed | llmrun Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.6 35B A3BAlibabaHow to run → Q4_K_M · 22.0 GB · 66.3 t/s | 66.3 tok/s | Q4_K_M · 22.0 GB | |
| 2 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB · 72.1 t/s | 72.1 tok/s | Q4_K_M · 16.5 GB | |
| 3 | Qwen3 30B A3B Thinking 2507Alibaba Q4_K_M · 19.4 GB · 84.9 t/s | 84.9 tok/s | Q4_K_M · 19.4 GB | |
| 4 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 77.9 t/s | 77.9 tok/s | Q4_K_M · 13.1 GB | |
| 5 | Q4_K_M · 19.3 GB · 108 t/s | 108 tok/s | Q4_K_M · 19.3 GB |
| # | Model | Speed | Coding Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.6 35B A3BAlibabaHow to run → Q4_K_M · 22.0 GB · 66.3 t/s | 66.3 tok/s | Q4_K_M · 22.0 GB | |
| 2 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB · 72.1 t/s | 72.1 tok/s | Q4_K_M · 16.5 GB | |
| 3 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 77.9 t/s | 77.9 tok/s | Q4_K_M · 13.1 GB | |
| 4 | Q4_K_M · 7.9 GB · 42.4 t/s | 42.4 tok/s | Q4_K_M · 7.9 GB | |
| 5 | Q4_K_M · 19.3 GB · 108 t/s | 108 tok/s | Q4_K_M · 19.3 GB |
| # | Model | Speed | Reasoning Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.6 35B A3BAlibabaHow to run → Q4_K_M · 22.0 GB · 66.3 t/s | 66.3 tok/s | Q4_K_M · 22.0 GB | |
| 2 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB · 72.1 t/s | 72.1 tok/s | Q4_K_M · 16.5 GB | |
| 3 | Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB · 35.4 t/s | 35.4 tok/s | Q4_K_M · 6.6 GB | |
| 4 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 77.9 t/s | 77.9 tok/s | Q4_K_M · 13.1 GB | |
| 5 | Q4_K_M · 19.3 GB · 108 t/s | 108 tok/s | Q4_K_M · 19.3 GB |
| # | Model | Speed | Math Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.6 35B A3BAlibabaHow to run → Q4_K_M · 22.0 GB · 66.3 t/s | 66.3 tok/s | Q4_K_M · 22.0 GB | |
| 2 | Qwen3.8 27BAlibaba Q4_K_M · 17.9 GB · 12.9 t/s | 12.9 tok/s | Q4_K_M · 17.9 GB | |
| 3 | Qwen3.6 27BAlibaba Q4_K_M · 17.9 GB · 12.9 t/s | 12.9 tok/s | Q4_K_M · 17.9 GB | |
| 4 | Q4_K_M · 19.3 GB · 108 t/s | 108 tok/s | Q4_K_M · 19.3 GB | |
| 5 | GLM 4.7 FlashZ.ai Q4_K_M · 19.9 GB · 64.1 t/s | 64.1 tok/s | Q4_K_M · 19.9 GB |
| # | Model | Speed | Long context Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Q4_K_M · 16.5 GB · 72.1 t/s | 72.1 tok/s | Q4_K_M · 16.5 GB | |
| 2 | Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB · 35.4 t/s | 35.4 tok/s | Q4_K_M · 6.6 GB | |
| 3 | Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB · 69.0 t/s | 69.0 tok/s | Q4_K_M · 3.4 GB | |
| 4 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 77.9 t/s | 77.9 tok/s | Q4_K_M · 13.1 GB | |
| 5 | LFM2.5 24B A2BLiquid AI Q4_K_M · 15.0 GB · 112 t/s | 112 tok/s | Q4_K_M · 15.0 GB |
| # | Model | Speed ↓ | llmrun Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Q4_K_M · 19.4 GB · 84.9 t/s | 84.9 tok/s | Q4_K_M · 19.4 GB | |
| 2 | Q4_K_M · 19.3 GB · 108 t/s | 108 tok/s | Q4_K_M · 19.3 GB | |
| 3 | Qwen3 30B A3B Instruct 2507Alibaba Q4_K_M · 19.4 GB · 84.9 t/s | 84.9 tok/s | Q4_K_M · 19.4 GB | |
| 4 | GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 77.9 t/s | 77.9 tok/s | Q4_K_M · 13.1 GB | |
| 5 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB · 72.1 t/s | 72.1 tok/s | Q4_K_M · 16.5 GB |
| # | Model | Speed | Vision Score | Quant · VRAM |
|---|---|---|---|---|
| 1 | Qwen3.6 35B A3BAlibabaHow to run → Q4_K_M · 22.0 GB · 66.3 t/s | 66.3 tok/s | Q4_K_M · 22.0 GB | |
| 2 | Gemma 4 26B A4B ITGoogle Q4_K_M · 16.5 GB · 72.1 t/s | 72.1 tok/s | Q4_K_M · 16.5 GB | |
| 3 | Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB · 35.4 t/s | 35.4 tok/s | Q4_K_M · 6.6 GB | |
| 4 | Gemma 3 12B ITGoogle Q4_K_M · 10.9 GB · 28.0 t/s | 28.0 tok/s | Q4_K_M · 10.9 GB | |
| 5 | MiniCPM V 4.7 35B A3BOpenBMB Q4_K_M · 22.1 GB · 76.4 t/s | 76.4 tok/s | Q4_K_M · 22.1 GB |
See all 327 models on the Tesla P40 →Best quant that fits · speeds estimated on this card
The Tesla P40 runs 14B and 35B MoE models well. Pascal generation, released 2016.
GenerationPascalreleased 2016
Tesla P40 strengths and weaknesses
Strengths
- Runs 35B MoE models: Qwen3.6 35B A3B at 66.3 tok/s in 22 GB
Weaknesses
- 70B models don't fit: Llama 3.3 70B Instruct needs 51.3 GB at Q4 vs 24 GB here
- Slow on large dense models: Qwen3.8 27B at 12.9 tok/s
- Slow prompt reading: A 4K-token prompt on Qwen3.6 35B A3B waits 7.3 s for the first token
- No FP8 or NVFP4: Pascal predates both; use GGUF, AWQ or GPTQ builds
Output speed depends on memory bandwidth (346 GB/s). Grey rows need more than 24 GB.
- 9BQwen3.5 9B · Q4_K_M · 6.6 GB35.4tok/s
- 14BQwen3 14B · Q4_K_M · 11.9 GB23.7tok/s
- 27–31BQwen3.8 27B · Q4_K_M · 17.9 GB12.9tok/s
- 35B MoEQwen3.6 35B A3B · Q4_K_M · 22.0 GB66.3tok/s
- 70BLlama 3.3 70B Instruct · not at 4-bitneeds 51.3 GB
- 120B MoEGPT OSS 120B · not at 4-bitneeds 70.8 GB
Estimated speeds: Q4_K_M, 4K context, one user. How we estimate
All 1141 models on the Tesla P40
Featured onlyBest quant that fits, ordered by llmrun Score. Fit grades how well it runs here, S best. How ranking works
| Model | Quant | VRAM | Speed | Context | Score | Fit |
|---|---|---|---|---|---|---|
Granite Vision 4.1 4BVisionFunctions | Q4_K_M | 4.0 GB | 78.4 t/s | 131K | S87, Runs great | |
NVIDIA Nemotron Labs Dory1.0 4B BF16ChatFunctions | Q4_K_M | 5.7 GB | 73.5 t/s | 262K | S86, Runs great | |
LocateAnything 3BVisionFunctions | Q4_K_M | 3.2 GB | 84.2 t/s | 33K | S88, Runs great | |
| Q4_K_M | 18.1 GB | 14.9 t/s | 131K | B55, Decent | ||
Qwen3.5 2BVisionFunctions | Q4_K_M | 1.9 GB | 127.1 t/s | 262K | S95, Runs great | |
Intern Decision 2BVisionFunctions | Q4_K_M | 1.8 GB | 130.0 t/s | 262K | S95, Runs great | |
WeVisOne VL 4BVisionFunctions | Q4_K_M | 4.7 GB | 66.3 t/s | 262K | S85, Runs great | |
HumanizerChat | Q4_K_M | 9.0 GB | 27.3 t/s | 262K | A68, Runs well | |
Clef FlashVisionFunctions | Q4_K_M | 6.5 GB | 36.2 t/s | 262K | A73, Runs well | |
| Q4_K_M | 16.9 GB | 13.6 t/s | 262K | B53, Decent | ||
| Q4_K_M | 1.5 GB | 190.6 t/s | 1049K | S99, Runs great | ||
| Q4_K_M | 2.6 GB | 227.2 t/s | — | S100, Runs great | ||
Kimi K3 DSparkChat | Q4_K_M | 2.2 GB | 130.8 t/s | 1049K | S95, Runs great | |
Holo4 35B A3BVisionFunctions | Q4_K_M | 21.5 GB | 78.1 t/s | 262K | A73, Runs well | |
Llm Jp 4.1 32B A3b ThinkingChatFunctions | Q4_K_M | 20.1 GB | 82.2 t/s | 66K | A82, Runs well |
Specs
- Architecture
- Pascal
- VRAM
- 24 GB GDDR5
- Memory bandwidth
- 346.0 GB/s
- FP16 (dense)
- 11.76 TFLOPS
- CUDA cores
- 3,840
- Tensor cores
- 0
- Compute capability
- 6.1
- TDP
- 250 W
- Released
- 2016-09-13
Tesla P40 FAQ
What is the largest model the Tesla P40 can run?
Qwen3.6 35B A3B (36.0B) is the largest at 4-bit or better: at Q4_K_M it uses 22.0 GB of 24 GB, and runs at about 66.3 tok/s.
Can the Tesla P40 run 70B models?
No. Llama 3.3 70B Instruct needs 51.3 GB at Q4, more than the 24 GB on this card.
How many tokens per second does the Tesla P40 give?
About 66.3 tok/s on Qwen3.6 35B A3B at Q4_K_M, with the first token after 7.3 s on a 4K-token prompt.
What power supply does the Tesla P40 need?
Plan a 500 W PSU (2× the 250 W TDP).
Is 24 GB enough for coding and agents?
Yes for coding: Qwen3.6 35B A3B, the best of the 4 coding models that fit, runs at about 66 tok/s. For agents, not yet: no agent-scored model fits in less than 96 GB.