Qwen3.8 — lokalny model AI od Alibaba Cloud
Autor: Jakub Rusinowski · Ostatnia aktualizacja:
Alibaba's August 2026 wave. The 27B is a natively multimodal dense model under Apache 2.0 with a 262K context that extends to 1M via YaRN — small enough for one consumer card, which is the combination most of this site's visitors are looking for. Qwen3.8-Max sits at the other extreme: 2.4T parameters with roughly 95B active, open-weighted under its own licence. Qwen3.8-Flash-Next is neither — an experimental preview of the Qwen4 architecture, 180B resident but only 6B active, under the Qwen Community License rather than Apache.
Warianty
Najmniejszy wariant Qwen3.8 potrzebuje około 18 GB VRAM przy Q4_K_M — skwantyzowane wagi plus narzut frameworka, bez pamięci podręcznej KV.
| Model | VRAM przy Q4 | VRAM | Kontekst | Uruchom |
|---|---|---|---|---|
| Qwen3.8 27B → 27B | ~17.6 GB | 262,144 | ollama run qwen3.8:27b | |
| Qwen3.8-Max → 2.4T/95B active | ~1449.8 GB | 262,144 | ollama run qwen3-8 | |
| Qwen3.8-Flash-Next → 180B (6B active) | ~109.5 GB | 262,144 | ollama run qwen3-8 |
Pamięć to skwantyzowane wagi plus narzut przy Q4_K_M, liczone tym samym silnikiem co sprawdzarka GPU i VRAM.
Jak uruchomić Qwen3.8 lokalnie
Zainstaluj Ollama, a potem pobierz tag.
ollama run qwen3.8:27bWybierz rozmiar powyżej, aby zobaczyć jego VRAM, szacowaną szybkość i polecenie instalacji.
Licencja
Commercial use permitted. No usage restrictions beyond attribution.
Dotyczy: Qwen3.8 27BWeights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.
Dotyczy: Qwen3.8-MaxWeights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.
Dotyczy: Qwen3.8-Flash-NextPolecane GPU
Najtańsze GPU z naszego katalogu, które uruchomi Qwen3.8 lokalnie (min. 18 GB VRAM), to AMD Radeon RX 7900 XT (20 GB).
Czy uruchomię Qwen3.8 na moim GPU?
- Qwen3.8 na AMD Radeon RX 7800 XT
- Qwen3.8 na AMD Radeon RX 7900 XT
- Qwen3.8 na AMD Radeon RX 7900 XTX
- Qwen3.8 na AMD Radeon RX 9060 XT 16GB
- Qwen3.8 na AMD Radeon RX 9060 XT 8GB
- Qwen3.8 na AMD Radeon RX 9070
- Qwen3.8 na AMD Radeon RX 9070 XT
- Qwen3.8 na Apple M1
- Qwen3.8 na Apple M1 Pro
- Qwen3.8 na Apple M2
- Qwen3.8 na Apple M2 Pro
- Qwen3.8 na Apple M2 Ultra
- Qwen3.8 na Apple M3
- Qwen3.8 na Apple M3 Pro
- Qwen3.8 na Apple M4
- Qwen3.8 na Apple M4 Pro
- Qwen3.8 na Apple M5
- Qwen3.8 na Intel Arc B570
- Qwen3.8 na Intel Arc B580
- Qwen3.8 na NVIDIA DGX Spark
- Qwen3.8 na NVIDIA GeForce RTX 3060 (12GB)
- Qwen3.8 na NVIDIA GeForce RTX 3070
- Qwen3.8 na NVIDIA GeForce RTX 3070 Ti
- Qwen3.8 na NVIDIA GeForce RTX 3080 (10GB)
- Qwen3.8 na NVIDIA GeForce RTX 3090
- Qwen3.8 na NVIDIA GeForce RTX 4060
- Qwen3.8 na NVIDIA GeForce RTX 4060 Ti 16GB
- Qwen3.8 na NVIDIA GeForce RTX 4070
- Qwen3.8 na NVIDIA GeForce RTX 4070 Super
- Qwen3.8 na NVIDIA GeForce RTX 4070 Ti
- Qwen3.8 na NVIDIA GeForce RTX 4070 Ti Super
- Qwen3.8 na NVIDIA GeForce RTX 4080
- Qwen3.8 na NVIDIA GeForce RTX 4080 Super
- Qwen3.8 na NVIDIA GeForce RTX 4090
- Qwen3.8 na NVIDIA GeForce RTX 5060
- Qwen3.8 na NVIDIA GeForce RTX 5060 Ti 16GB
- Qwen3.8 na NVIDIA GeForce RTX 5060 Ti 8GB
- Qwen3.8 na NVIDIA GeForce RTX 5070
- Qwen3.8 na NVIDIA GeForce RTX 5070 Ti
- Qwen3.8 na NVIDIA GeForce RTX 5080
- Qwen3.8 na NVIDIA GeForce RTX 5090
Qwen3.8 — najczęściej zadawane pytania
How much VRAM does Qwen3.8 need?
Qwen3.8 needs about 18 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Qwen3.8 27B (18 GB, Q4_K_M); Qwen3.8-Max (1450 GB, Q4_K_M); Qwen3.8-Flash-Next (109 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Qwen3.8 on an RTX 4090 (24 GB)?
Yes — Qwen3.8 runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for Qwen3.8?
Q4_K_M is the best balance of quality and VRAM for Qwen3.8 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Qwen3.8 with Ollama?
Install Ollama, then run: ollama run qwen3.8:27b. This downloads Qwen3.8 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.