Mistral-Nemo-Instruct-2407 · 11 GPUs measured first-party · llama.cpp · Updated October 2026
What GPU Do You Need for Mistral-Nemo-Instruct-2407?
Mistral-Nemo-Instruct-2407 on 11 GPUs, measured first-party: NVIDIA B300 leads at 196.9 tok/s, T4 trails at 23 tok/s, and it peaked at 8GB of VRAM.
Benchmarked weights: bartowski/Mistral-Nemo-Instruct-2407-GGUF
NVIDIA B300
196.9 tok/s on Mistral-Nemo-Instruct-2407, the ceiling. Measured on our bench. 288GB of VRAM, $40,000 at launch.
- 196.9 tok/s on Mistral-Nemo-Instruct-2407
- 288GB, clears the Mistral-Nemo-Instruct-2407 floor
- Rentable by the hour rather than bought
- 1400W board rating
- Datacenter or workstation hardware, not a retail purchase
NVIDIA RTX PRO 6000 Blackwell Workstation Edition
164.3 tok/s on Mistral-Nemo-Instruct-2407, lowest launch price that still fits. Measured on our bench. 96GB of VRAM, $8,565 at launch.
- 164.3 tok/s on Mistral-Nemo-Instruct-2407
- 96GB, clears the Mistral-Nemo-Instruct-2407 floor
- Rentable by the hour rather than bought
- 600W board rating
- Datacenter or workstation hardware, not a retail purchase
What GPU Do You Need for Mistral-Nemo-Instruct-2407?, tok/s by GPU
Efficiency: tok/s per 100W drawn
Power is the average pulled during the run, sampled at 1Hz. The fastest card is often not the one here, and for anything left running this is the number that shows up on the bill.
Value: tok/s per $1,000 of MSRP
Launch price, not street price, so it ages. A speed leaderboard always crowns the most expensive card; this is the counterweight.
Mistral-Nemo-Instruct-2407. Measured tokens per second by GPU
| GPU | tok/s | Prompt t/s | tok/W | Avg power |
|---|---|---|---|---|
| NVIDIA B300 | 196.9 | 3753.8 | 0.48 | 406.1 W |
| NVIDIA B200 | 183.9 | 6668.3 | 0.43 | 426.5 W |
| NVIDIA H200 | 181.3 | 5996.2 | 0.69 | 264.3 W |
| NVIDIA H100 80GB HBM3 | 177.4 | 6042.3 | 0.63 | 280.4 W |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | 164.3 | 9002.4 | 0.63 | 262.2 W |
| NVIDIA A100 80GB SXM4 | 111.1 | 3016.7 | 0.48 | 229.9 W |
| NVIDIA A100 40GB SXM4 | 106.3 | 3001.5 | 0.62 | 170.8 W |
| NVIDIA L40S | 89.72 | 6597.4 | 0.41 | 218.4 W |
| NVIDIA A10G | 57 | 2082.8 | 0.46 | 125.1 W |
| NVIDIA L4 | 33.01 | 1927.1 | 0.52 | 62.9 W |
| NVIDIA T4 | 22.98 | 804 | 0.36 | 63.8 W |
What the numbers show. Across 11 GPUs measured on our own bench, B300 is fastest at 197 tok/s. The slowest, T4, manages 23.0, so the spread is 8.6x from top to bottom. H200 is the most efficient, 181 tok/s at 264W. Per dollar of launch price, A10G gives the most (20.4 tok/s per $1,000). The fastest card with 16GB or less is T4 at 23.0 tok/s.
How it compares. H100 80GB HBM3: Mistral-Nemo-Instruct-2407 177.4 tok/s, NemoMix-Unleashed-12B 177.3, gemma-2-9b 174.7 (9B), GLM-4.7-Flash 186.1 (31B), Qwen3-Next-80B-A3B 186.2. 2 of 4 beat Mistral-Nemo-Instruct-2407 here. Qwen3-Next-80B-A3B is a mixture-of-experts, so per token it computes only a slice of its size.
Cost on a rented GPU. 1M generated tokens of Mistral-Nemo-Instruct-2407: $1.23 on a A100 40GB SXM4 ($0.47/hr, 2.6 hours), $9.79 on a B300 ($6.94/hr, 85 min, 7.9x the cost).
Mistral-Nemo-Instruct-2407: cost per 1M generated tokens on rented GPUs
| GPU | Cheapest rate | Speed (tok/s) | Cost per 1M generated tokens |
|---|---|---|---|
| NVIDIA A100 40GB SXM4 | $0.47/hr | 106.3 | $1.23 |
| NVIDIA T4 | $0.14/hr | 22.98 | $1.64 |
| NVIDIA RTX PRO 6000 Blackwell Workstation Edition | $1.08/hr | 164.3 | $1.82 |
| NVIDIA A100 80GB SXM4 | $0.95/hr | 111.1 | $2.37 |
| NVIDIA L40S | $0.79/hr | 89.72 | $2.45 |
| NVIDIA H100 80GB HBM3 | $2.14/hr | 177.4 | $3.34 |
| NVIDIA L4 | $0.44/hr | 33.01 | $3.70 |
| NVIDIA H200 | $3.59/hr | 181.3 | $5.50 |
| NVIDIA B200 | $5.98/hr | 183.9 | $9.03 |
| NVIDIA B300 | $6.94/hr | 196.9 | $9.79 |
Cheapest hourly rate we track on RunPod and Vast.ai, divided by the measured speed. Startup time and storage are extra.
Speed tiers for Mistral-Nemo-Instruct-2407. 30+ tok/s: 10 (B300, B200, H200); 10-30 tok/s: 1 (T4). 30 tok/s is roughly where replies outpace reading.
Reading your prompt. Before Mistral-Nemo-Instruct-2407 writes anything it reads the input: 9002.4 tok/s on the RTX PRO 6000 Blackwell Workstation Edition (0.4s for a 4,000-token prompt), 804.0 on the T4 (5.0s). Long documents and big code files feel this number more than the generation speed.
VRAM for Mistral-Nemo-Instruct-2407. Measured peak 7.5GB, so 8GB is the smallest common card size; smallest card it ran on: T4 (16GB). With long context: Q4_K_M 10GB (tested), Q2_K 7GB, Q3_K_M 9GB, Q5_K_M 12GB, Q6_K 14GB.
Power on Mistral-Nemo-Instruct-2407. Most efficient: H200, 264W, 0.40 kWh per 1M generated tokens. Hungriest: B200, 426W, 0.64 kWh. At $0.15/kWh: $0.061 per 1M generated tokens.
Our verdict
Fastest on Mistral-Nemo-Instruct-2407: NVIDIA B300, 196.9 tok/s. Cheapest to rent per job: NVIDIA A100 40GB SXM4, $1.23 per 1M generated tokens.
FAQ
- What GPU do I need to run Mistral-Nemo-Instruct-2407?
- About 8GB. Smallest card that ran it: NVIDIA T4 (16GB).
- How much does it cost to run Mistral-Nemo-Instruct-2407 in the cloud?
- $1.23 per 1M generated tokens on a NVIDIA A100 40GB SXM4 at $0.47/hr, cheapest of 10 rentable cards we measured.
- Can I run Mistral-Nemo-Instruct-2407 on a 12GB, 16GB or 24GB card?
- It used 7.5GB at the precision we tested. 12GB: yes; 16GB: yes; 24GB: yes.
- Is the H100 80GB HBM3 or the A100 80GB SXM4 faster for Mistral-Nemo-Instruct-2407?
- The H100 80GB HBM3: 177.4 vs 111.1 tok/s, 60% faster on our bench.
How we test
llama.cpp llama-bench at Q4_K_M, 512-token prompt and 128 generated tokens, three runs after a warmup, full GPU offload, with power and VRAM sampled throughout. Token generation is memory-bandwidth-bound, so the ranking tracks bandwidth closely, which makes it a fair guide to cards we haven't run yet.