All compatibility checks

Compatibility check

Can you run DeepSeek R1 Distill Qwen 7B on the RTX 4070 SUPER?

Yes — with quantization

Yes, with quantization. DeepSeek R1 Distill Qwen 7B does not fit the RTX 4070 SUPER (12 GB) in FP16, but it runs at INT8 (8-bit) using about 9 GB (75% of VRAM). Use a GPTQ, AWQ, or GGUF build to get there, and keep prompts moderate to leave room for the KV cache.

Memory breakdown

Weights plus a 0.2 GB KV cache at 4,096tokens, against the card's 12 GB. Verdicts leave ~10% headroom for activations and fragmentation.

PrecisionWeightsKV cacheTotal% of 12 GBFit
FP16 / BF16full quality17.5 GB0.2 GB17.7 GB148%No
INT8 (8-bit)near-full quality8.8 GB0.2 GB9 GB75%Fits
INT4 (4-bit)GPTQ / AWQ / GGUF Q44.4 GB0.2 GB4.6 GB38%Fits

Planning estimates, not a substitute for profiling. Real usage varies with the inference runtime, batch size, and how much context you actually use — the KV cache grows linearly with prompt length.

GPUs that run DeepSeek R1 Distill Qwen 7B

Cards where this model fits (at its best precision):

Go deeper

Frequently asked questions

Can the RTX 4070 SUPER run DeepSeek R1 Distill Qwen 7B?

Not in FP16, but yes at INT8 (8-bit), where it uses about 9 GB versus the card's 12 GB.

How much VRAM does DeepSeek R1 Distill Qwen 7B need?

Approximately 17.5 GB in FP16, 8.8 GB in INT8, and 4.4 GB in 4-bit for the weights, plus a KV cache of about 0.2 GB at 4,096 tokens.

Does quantization let DeepSeek R1 Distill Qwen 7B fit on the RTX 4070 SUPER?

Yes. Dropping to INT8 (8-bit) brings total usage to about 9 GB, which fits the 12 GB card with headroom for the KV cache.

What happens to memory with longer context?

The KV cache grows linearly with prompt length. At 4,096 tokens it is about 0.2 GB here; doubling the context roughly doubles that term, so long-context use can push a tight fit over the edge.