GLM-5.3 vs DeepSeek V4.1 Flash
Too close to call. On the work we tested, GLM-5.3 and DeepSeek V4.1 Flash can't be told apart. DeepSeek V4.1 Flash is the cheaper one, by 16×.
GrindBench-PubSec
BGLM-5.3Z.ai (Zhipu)68% 63–74%8.5¢ a task · 300 tasksBDeepSeek V4.1 FlashDeepSeek64% 59–70%0.43¢ a task · 300 tasks
Too close to call here: their ranges overlap, so on another set of tasks the order could flip.
| Kind of work | GLM-5.3 | DeepSeek V4.1 Flash |
|---|---|---|
| ATIP redaction | 16% | 20% |
| Procurement evaluation | 88% | 78% |
| Performance analysis | 69% | 63% |
| Bilingual equivalence | 95% | 95% |
GrindBench-Banking
BGLM-5.3Z.ai (Zhipu)70% 58–82%$0.23 a task · 60 tasksBDeepSeek V4.1 FlashDeepSeek62% 48–73%1.6¢ a task · 60 tasks
Too close to call here: their ranges overlap, so on another set of tasks the order could flip.
| Kind of work | GLM-5.3 | DeepSeek V4.1 Flash |
|---|---|---|
| Lending & mortgages | 60% | 60% |
| AML & KYC | 60% | 70% |
| Deposits & payments | 60% | 40% |
| Complaints & conduct | 90% | 70% |
| Credit risk | 80% | 80% |
| Capital & liquidity | 70% | 50% |
Each kind of work here rests on 10 tasks per model, so a single row can easily be 20 points off. A lead of 10 points or more is in bold; read it as a hint, not a result.