← Back to models

Qwen3-235B-A22B

Alibaba · Mixture-of-Experts (MoE) · License: Apache-2.0

Alibaba's flagship Qwen3 MoE model. Unlike several earlier Qwen releases (which use a custom license), this one ships under a standard Apache-2.0 license.

Memory requirement — the visual calculator

In real deployments, memory needs also depend on numeric precision (e.g. FP16 vs INT8 vs INT4), which changes bytes-per-parameter. This store uses a simplified, transparent formula so beginners can follow the logic: (parameters in billions × 1 GB) + 20% overhead for the runtime, active requests, and context memory.

Model weights
235 GB
+ 20% overhead
47.0 GB
= Total required
282.0 GB

Calculation: 235B parameters × 1 GB = 235 GB, plus 20% (47.0 GB) = 282.0 GB total The dedicated memory built into the graphics card. An AI model's entire set of "weights" (its learned parameters) must fit into this memory before the card can run it. If a model needs more VRAM than one card has, you need multiple cards working together, or a card with more memory. needed. Note: even though only 22B parameters ( A model design where only a fraction of the model's parameters ("active parameters") are used to process any single piece of text, even though the full model is much larger. This makes the model faster to run than its total size suggests, but you still need enough memory to hold the entire model, not just the active part. ) are active per token, the full 235B must still be loaded into memory.

Cheapest way to run this model today
3× NVIDIA RTX PRO 6000 Blackwell (Workstation Edition) (288 GB total VRAM, ~$34,908). See the full recommendation →

Leaderboard

Added to the LMArena text leaderboard in August 2025.

Sources