IBM vs 
Mistral AI
Verdict
- IBM's flagship leads: Granite 4.2 30B at 50.0 vs Mistral Medium 3.5 at 49.0.
- IBM has the better model in 4 of 4 price bands.
- IBM's fastest ranked model (Granite 4.2 3B, 214 tok/s) beats Mistral AI's (Ministral 3 3B, 202 tok/s).
- Mistral AI publishes more open-weights models (38 vs 10).
- Mistral AI ships more often: one model every 46 days against every 118.
Best model per price band
Each provider’s highest-index model below the price ceiling. Purple marks the winner of the band.
| Price band | IBM | Mistral AI |
|---|---|---|
| Under $0.50/M | Granite 4.2 30B 50.0$0.283/M | Mistral Small 4 44.8$0.262/M |
| Under $2/M | Granite 4.2 30B 50.0$0.283/M | Mistral Large 3 45.0$0.750/M |
| Under $10/M | Granite 4.2 30B 50.0$0.283/M | Mistral Medium 3.5 49.0$3.00/M |
| Any price | Granite 4.2 30B 50.0$0.283/M | Mistral Medium 3.5 49.0$3.00/M |
Line-up
| Metric | IBM | Mistral AI |
|---|---|---|
| Flagship index | 50.0 | 49.0 |
| Ranked models | 7 | 24 |
| Open-weights models | 10 | 38 |
| Cheapest ranked, $/M | $0.053 | $0.100 |
| Fastest ranked, tok/s | 214 | 202 |
| Release cadence, days | 118 | 46 |
| Last release | 25 Aug 2026 | 13 Jun 2026 |
| Next expected | 3 Nov 2026 – 6 Feb 2027 | 10 Jul 2026 – 16 Aug 2026 |
Data as of 13 Sept 2026. See IBM and Mistral AI in full.