Qwen3 8B
Qwen3 8B needs 6.7 GB of VRAM at Q4_K_M with 8k context, computed from its config.json as checked on 16 October 2026. The smallest card that holds it has 8 GB, and the fastest desktop card that runs it, the RTX 5090, generates at most 264 tokens per second.
- Parameters
- 8.2B
- Context
- 40,960 tokens
- Licence
- Apache 2.0
- Released
- April 2025
- Config checked
- 16 October 2026
Open Qwen3 8B in the LLM checker
Which card size runs it
6.7 GB against each memory size. A model may use 95% of a card, and only what is left over, or missing, is coloured; tight means above 85%, with little room for a longer conversation.
Other quantisations, context lengths or a Q8 KV cache: the VRAM calculator.
Computed 16 October 2026Which GPU to pick
The smallest card here that loads it at Q4_K_M with 8k of context is the RTX 3060 (12 GB): 6.7 GB needed, an estimated 53 tokens per second. That is only the smallest card with its own page here: the model itself fits any 8 GB card. For room to work, meaning 32,768 tokens of context while staying under 85% of memory, step up to the RTX 4060 Ti (16 GB), which needs 10.3 GB for that and generates at about 42 tokens per second.
The same card also holds it at Q8_0, in 10.7 GB. Q8 is near-lossless, so take the higher precision when it costs nothing. On Apple silicon, a Mac mini M4 Pro holds it with 32,768 tokens of context at an estimated 40 tokens per second; prompt processing is slower than on Nvidia, which long documents make obvious. Among unified-memory desktops, the Ryzen AI Max+ 395 holds it at 32,768 tokens in 10.3 GB, at an estimated 38 tokens per second.
Speed on each card
| GPU | Fits | Est. tokens/s | Feels like | Memory | Bandwidth |
|---|---|---|---|---|---|
| H100 SXM | Yes | ≤ 493 | faster than you can read | 80 GB | 3,350 GB/s |
| A100 | Yes | ≤ 300 | faster than you can read | 80 GB | 2,039 GB/s |
| RTX 5090 | Yes | ≤ 264 | faster than you can read | 32 GB | 1,792 GB/s |
| RTX PRO 6000 Blackwell | Yes | ≤ 264 | faster than you can read | 96 GB | 1,792 GB/s |
| RTX PRO 5000 Blackwell | Yes | ≤ 198 | faster than you can read | 48 GB | 1,344 GB/s |
| RTX 4090 | Yes | ≤ 148 | faster than you can read | 24 GB | 1,008 GB/s |
| RTX 5080 | Yes | ≤ 141 | faster than you can read | 16 GB | 960 GB/s |
| RX 7900 XTX | Yes | ≤ 141 | faster than you can read | 24 GB | 960 GB/s |
| RTX 6000 Ada | Yes | ≤ 141 | faster than you can read | 48 GB | 960 GB/s |
| RTX 3090 | Yes | ≤ 138 | faster than you can read | 24 GB | 936 GB/s |
| RTX 5070 Ti | Yes | ≤ 132 | faster than you can read | 16 GB | 896 GB/s |
| RTX PRO 4500 Blackwell | Yes | ≤ 132 | faster than you can read | 32 GB | 896 GB/s |
| L40S | Yes | ≤ 127 | faster than you can read | 48 GB | 864 GB/s |
| RX 7900 XT | Yes | ≤ 118 | faster than you can read | 20 GB | 800 GB/s |
| RTX A6000 | Yes | ≤ 113 | faster than you can read | 48 GB | 768 GB/s |
| RTX 4080 SUPER | Yes | ≤ 108 | faster than you can read | 16 GB | 736 GB/s |
| RTX 5070 | Yes | 99 | faster than you can read | 12 GB | 672 GB/s |
| RTX PRO 4000 Blackwell | Yes | 99 | faster than you can read | 24 GB | 672 GB/s |
| RX 9070 XT | Yes | 94 | faster than you can read | 16 GB | 640 GB/s |
| Radeon AI PRO R9700 | Yes | 94 | faster than you can read | 32 GB | 640 GB/s |
| Arc Pro B70 | Yes | 89 | faster than you can read | 32 GB | 608 GB/s |
| RTX 4070 | Yes | 74 | faster than you can read | 12 GB | 504 GB/s |
| Arc B580 | Yes | 67 | faster than you can read | 12 GB | 456 GB/s |
| Arc Pro B60 | Yes | 67 | faster than you can read | 24 GB | 456 GB/s |
| RTX 5060 Ti | Yes | 66 | faster than you can read | 16 GB | 448 GB/s |
| RTX PRO 4000 Blackwell SFF | Yes | 64 | faster than you can read | 24 GB | 432 GB/s |
| RTX 3060 | Yes | 53 | comfortable for chat | 12 GB | 360 GB/s |
| RX 9060 XT | Yes | 47 | comfortable for chat | 16 GB | 320 GB/s |
| RTX Spark | Yes | 44 | comfortable for chat | 128 GB (102 for the GPU) | 300 GB/s |
| RTX 4060 Ti | Yes | 42 | comfortable for chat | 16 GB | 288 GB/s |
| DGX Spark | Yes | 40 | comfortable for chat | 128 GB (126 for the GPU) | 273 GB/s |
| Ryzen AI Max+ 395 | Yes | 38 | comfortable for chat | 128 GB (96 for the GPU) | 256 GB/s |
Single-stream decode, from memory bandwidth at 70% efficiency. Prompt processing and batching not included. These are ceilings, not measurements: Qwen3 8B reads only 4.8 GB of weights per token at Q4, so little that the per-token costs the formula leaves out decide much of the real speed. The 16 figures marked ≤ are where a real runtime falls furthest below the number shown. See the speed estimator for other quantisations and Apple hardware, or every card compared if you are choosing hardware rather than a model.
Computed 16 October 2026Memory by quantisation and context
| Quant | 4,096 ctx | 8,192 ctx | 32,768 ctx |
|---|---|---|---|
| Q8_0 | 10.1 GB · fits 12 GB | 10.7 GB · fits 12 GB | 14.4 GB · fits 16 GB |
| Q5_K_M | 7.2 GB · fits 8 GB | 7.8 GB · fits 12 GB | 11.4 GB · fits 12 GB |
| Q4_K_M | 6.1 GB · fits 8 GB | 6.7 GB · fits 8 GB | 10.3 GB · fits 12 GB |
Qwen3.5 9B is the newer model in this family. It needs about 6.7 GB at the same setting.
Weights at this quant: 4.8 GB at Q4. Every extra 1,000 tokens of context adds about 0.147 GB of KV cache at FP16. Try other settings in the VRAM calculator.
About the model
Qwen3 8B has 8.2B parameters, 36 layers and 8 KV heads of dimension 128. The 8B dense model of Alibaba's Qwen3 generation, April 2025. A standard transformer with a thinking switch, and the model a great many 2025 local setups were built around.
Text only 40,960-token window Apache 2.0 April 2025
Reasoning. Switchable: it can think step by step before answering or reply directly, set per request.
Native context window: 40,960 tokens. Licence: Apache 2.0, a permissive licence. Weights published in April 2025.
Good at
- General chat and light coding on an 8 to 12 GB card.
- Setups that depend on Qwen3-era fine-tunes and adapters.
Watch out for
- The native window is 40,960 tokens. Qwen documents YaRN scaling to reach 131k, but that is a stretch of a 2025 design, not what it was trained at.
- Every layer keeps a full-length cache, so long context costs several times what a 2026 hybrid model pays for the same conversation.
- Qwen3.5 9B replaces it, adds vision, and holds six times the context natively.
Similar models
The models that need about the same memory as Qwen3 8B at Q4_K_M and 8k, which makes them the real alternatives on whatever card you have. The context column is what each extra 1,000 tokens costs, and it is where models of the same size differ most.
| Model | Size | Needs | Per 1k context | Window | Released |
|---|---|---|---|---|---|
| Qwen3 8B | 8.2B | 6.7 GB | 0.147 GB | 40k | 2025-04 |
| Qwen3.5 9B | 9.7B | 6.7 GB | 0.033 GB | 256k | 2026-02 |
| Llama 3.1 8B | 8B | 6.4 GB | 0.131 GB | 128k | 2024-07 |
| Granite 4.2 8B | 8.8B | 7.2 GB | 0.164 GB | 128k | 2026-08 |
| Gemma 4 12B | 12B | 8.2 GB | 0.016 GB | 256k | 2026-05 |
See every model compared, which models fit on 8 to 96 GB, or the cheapest hour of every cloud card.
Questions
- Can I run Qwen3 8B on a 24 GB card?
- Yes, at Q4_K_M and 8k context it needs about 6.7 GB, which fits a 24 GB card with room.
- How much VRAM does Qwen3 8B need at Q8?
- About 10.7 GB at 8k context, or 14.4 GB at 32k. Q8 is near-lossless; use it when it fits.
- What is the context length of Qwen3 8B?
- 40,960 tokens natively. That is short by current standards, and it is a limit of the model rather than of your GPU: spare memory does not extend it. Qwen documents YaRN scaling to stretch it further, at some cost in quality on short prompts.
- Is there a newer model than Qwen3 8B?
- Yes: Qwen3.5 9B, published in February 2026. At the same Q4_K_M and 8k setting it needs about 6.7 GB against 6.7 GB for this one.
- How fast is Qwen3 8B on an RTX 4090?
- At most about 148 tokens per second at Q4, single stream, which is faster than you can read. That is a ceiling worked out from the card's memory bandwidth, not a measurement, and at a figure this high real runtimes land well below it, because the costs the formula leaves out take a growing share of each token.
- Architecture values from config.json in Qwen/Qwen3-8B on Hugging Face, checked 16 October 2026. Every figure on this page is computed from those values; check them before buying hardware for this model.
- Estimates assume a single conversation on a card that is otherwise free. A desktop on the same GPU takes 0.5 to 2 GB.