Magistral Small vs Qwen3 VL 30B A3B Thinking.
model vs model · prices as of
Qwen3 VL 30B A3B Thinking takes 2 of 5 rounds against Magistral Small. Both cost $1.05 per million tokens to run.
Mistral AIMagistral Small
$1.05/M tokens
cheapest we can explain, at Eden AI · $0.126 per 1,000 pages
128k context · tool calling · open weights
Alibaba Cloudwins 2 roundsQwen3 VL 30B A3B Thinking
$1.05/M tokens
cheapest we can explain, at Fireworks AI · $0.126 per 1,000 pages
262k context · tool calling · reads images · open weights
List price in (lower wins)
$0.50/M
$0.20/M
List price out (lower wins)
$1.50/M
$2.40/M
Cheapest we can explain (lower wins)
$1.05/M at Eden AI
$1.05/M at Fireworks AI
Standard price (lower wins)
$3/M
$3/M
Cost per 1,000 pages (lower wins)
$0.126
$0.126
Context
128k tokens
262k tokens
Open weights
yes
yes
Tool calling
yes
yes
Structured output
not stated
yes
Reads images
no
yes
Reasoning
yes
yes
Released
17 Mar 2025
6 Oct 2025
Providers selling it
6
10
Price we measure against
the maker's own
the maker's own
Trains on your prompts (no is better)
not stated
not stated
Providers selling both · 1
Eden AI$1.05 / $1.95
Magistral Small then Qwen3 VL 30B A3B Thinking, per million tokens
Go further
Add a third or fourth model, or switch the weighting, in the tool.
What moved in the last seven days
- Qwen3 VL 30B A3B Thinking
Nous Portal$0.16 / $1.92 $0.20 / $2.4025% dearer ·
Frequently asked
- Which is cheaper, Magistral Small or Qwen3 VL 30B A3B Thinking?
- Neither: both run at $1.05 per million tokens, three parts input to one part output, read on 16 Sept 2026.
- Where is each one cheapest?
- Magistral Small: $1.05 at Eden AI. Qwen3 VL 30B A3B Thinking: $1.05 at Fireworks AI.
- Can I buy both from one provider?
- Yes. One provider sells both: Eden AI. The cheapest of them for the two together is Eden AI.
- Which holds more context?
- Qwen3 VL 30B A3B Thinking, at 262k tokens against 128k.
Prices exclude tax and are per million tokens, read 16 Sept 2026 from each provider's own catalog. The standard price blends three parts input to one part output; a page is 600 tokens in and 60 out. Bars are scaled to the larger value in each row.