DeepSeek-R1-Distill-Qwen-32B vs DeepSeek-R1-Distill-Llama-70B
DeepSeek-R1-Distill-Qwen-32B and DeepSeek-R1-Distill-Llama-70B both score 8 on the intelligence index. The cheapest machine here that runs DeepSeek-R1-Distill-Llama-70B is the Framework Desktop, 128GB, at $3,449. DeepSeek-R1-Distill-Qwen-32B runs on the Radeon AI PRO R9700, 32GB at $1,299, $2,150 less. The Radeon AI PRO R9700, 32GB is priced as the card alone, without the PC around it. On the Framework Desktop, 128GB, the cheapest machine here that runs both, DeepSeek-R1-Distill-Qwen-32B is about 1.7× quicker: 6.8 tok/s against 4, DeepSeek-R1-Distill-Llama-70B's measured and the other estimated from memory bandwidth. At 500k tokens a day the Framework Desktop, 128GB never pays for itself on either.
| DeepSeek-R1-Distill-Qwen-32B | DeepSeek-R1-Distill-Llama-70B | |
|---|---|---|
| Intelligence index | 8 | 8 |
| Class | Below every hosted tier | Below every hosted tier |
| Weights | 20 GB | 43 GB |
| Needs at 32k | 28 GB | 53 GB |
| Quantisation | Q4_K_M | Q4_K_M |
| Parameters | 32.8B | 70.6B |
| Max context | 128k | 128k |
| API price per 1M | $0.018 in / $0.09 outpriced as gpt-oss-20b | $0.018 in / $0.09 outpriced as gpt-oss-20b |
| Licence | MIT | MIT |
| Machines here that run it | 26 of 37 | 13 of 37 |
| Cheapest machine that runs it | AMD Radeon AI PRO R9700, 32GB $1,299card only | Strix Halo Framework Desktop, 128GB $3,449 |
| Summarising | usable | usable |
| Translation | usable | usable |
| Everyday coding | usable | usable |
| Reasoning & maths | good | good |
| Agentic work | don’t | don’t |
Run DeepSeek-R1-Distill-Qwen-32B on the Radeon AI PRO R9700, 32GB · or DeepSeek-R1-Distill-Llama-70B on the Framework Desktop, 128GB
Ratings are coarse on purpose: they say what a model is usable for, not where it places to the decimal.
Jump to: Side by side on the Framework Desktop, 128GB · How much use it takes to pay for the machine · Machines that run one and not the other · The assumptions behind both columns
Side by side on the Framework Desktop, 128GB
The table above gives each model the cheapest machine that runs it, and those are two different machines, so nothing in it is a like-for-like race. The Strix Halo Framework Desktop, 128GB is the cheapest machine here that runs both, so this is the pair doing the same work on the same hardware.
| DeepSeek-R1-Distill-Qwen-32B | DeepSeek-R1-Distill-Llama-70B | |
|---|---|---|
| Speed at 32k | 6.8 tok/s estimated | 4 tok/s measured |
| Pay-back on this machine | Never pays back | Never pays back |
| API cost per month | $0.34priced as gpt-oss-20b | $0.34priced as gpt-oss-20b |
Run DeepSeek-R1-Distill-Qwen-32B on the Framework Desktop, 128GB · or DeepSeek-R1-Distill-Llama-70B
How much use it takes to pay for the machine
Everything above is at 500k tokens a day. Pay-back moves with how much you actually run, so here are both models at the five levels of use the calculator names, on the Framework Desktop, 128GB. Neither pays for it at any of those levels, even with agents running most of the day.
| A day's use | DeepSeek-R1-Distill-Qwen-32B | DeepSeek-R1-Distill-Llama-70B |
|---|---|---|
| 50ka few chats a day | Never pays back | Never pays back |
| 200klight assistant use | Never pays back | Never pays back |
| 1Ma moderate coding-assistant day | Never pays back | Never pays back |
| 4Mheavy coding with an agent | Never pays back | Never pays back |
| 20Magents running most of the day | Never pays backits ceiling | Never pays backits ceiling |
The Framework Desktop, 128GB cannot generate 20M tokens a day on either model: at most 9.33M on DeepSeek-R1-Distill-Qwen-32B and 5.49M on DeepSeek-R1-Distill-Llama-70B. Both figures on that row are for the most it can do.
Run DeepSeek-R1-Distill-Qwen-32B at 20M tokens a day · or DeepSeek-R1-Distill-Llama-70B
Machines that run one and not the other
DeepSeek-R1-Distill-Qwen-32B needs 28 GB of memory at 32k of context and DeepSeek-R1-Distill-Llama-70B needs 53 GB. That puts DeepSeek-R1-Distill-Qwen-32B on 13 of the 37 machines priced here that DeepSeek-R1-Distill-Llama-70B does not, starting at $1,299.
| Machine | Price | Memory | Speed on DeepSeek-R1-Distill-Qwen-32B | Pay-back |
|---|---|---|---|---|
| AMD Radeon AI PRO R9700, 32GB | $1,299card only | 32 GB | 20 tok/s estimated | Never pays back |
| Strix Halo Corsair AI Workstation 300, 64GB | $1,700 | 64 GB | 6.8 tok/s estimated | Never pays back |
| Strix Halo Framework Desktop, 64GB | $1,959 | 64 GB | 6.8 tok/s estimated | Never pays back |
| NVIDIA GeForce RTX 5090, 32GB | $1,999card only | 32 GB | 43 tok/s estimated | Never pays back |
| Strix Halo GMKtec EVO-X2, 64GB | $2,200 | 64 GB | 6.8 tok/s estimated | Never pays back |
| Mac mini M5 Pro, 48GB | $2,299 | 48 GB | 8.1 tok/s estimated | Never pays back |
7 more, on the DeepSeek-R1-Distill-Qwen-32B page.
DeepSeek-R1-Distill-Qwen-32B is also head to head with Devstral 2 123B above it on the leaderboard. One more model needs much the same memory: Qwen3 32B. DeepSeek-R1-Distill-Llama-70B is also head to head with Llama 3.3 70B Instruct below it on the leaderboard. One more model needs much the same memory: Qwen3-Coder Next.
The assumptions behind both columns
Both columns use the same usage: 500k tokens a day at 15:1 input to output, 32k of context, $0.17 per kWh, and today's API prices held flat. Speeds marked estimated are worked out from memory bandwidth rather than measured, and pay-back scales with them. Where nobody rents an open model by the token, its API prices are the nearest hosted model's, named beside them. Machines are the 37 here with a published price that are still sold. Change any of it in the calculator.
More head to head: every machine that runs DeepSeek-R1-Distill-Qwen-32B · every machine that runs DeepSeek-R1-Distill-Llama-70B · every other match-up · both against the frontier · the quickest pay-back at each level of use