GLM-5.3 vs GPT-6.1 Sol
Too close to call. On the work we tested, GLM-5.3 and GPT-6.1 Sol can't be told apart. GPT-6.1 Sol is the cheaper one, by 4.7×.
GrindBench-PubSec
BGLM-5.3Z.ai (Zhipu)68% 63–74%8.5¢ a task · 300 tasksCGPT-6.1 SolOpenAI58% 52–63%1.3¢ 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 | GPT-6.1 Sol |
|---|---|---|
| ATIP redaction | 16% | 0% |
| Procurement evaluation | 88% | 80% |
| Performance analysis | 69% | 60% |
| Bilingual equivalence | 95% | 85% |
GrindBench-Banking
BGLM-5.3Z.ai (Zhipu)70% 58–82%$0.23 a task · 60 tasksBGPT-6.1 SolOpenAI63% 50–75%5.4¢ 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 | GPT-6.1 Sol |
|---|---|---|
| Lending & mortgages | 60% | 60% |
| AML & KYC | 60% | 30% |
| Deposits & payments | 60% | 60% |
| Complaints & conduct | 90% | 100% |
| Credit risk | 80% | 60% |
| Capital & liquidity | 70% | 70% |
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.