Results
Same score. A fraction of the cost.
On all four COMPASS workflows, 360 ONE's model by Dvar matches GPT-6.1 Sol, the strongest frontier model we tested.
| COMPASS workflow | 360 ONE model · by Dvar | GPT-6.1 Sol | Cost per 10 lakh tasks | Cheaper |
|---|---|---|---|---|
| Client instructionsWealth Assistant · servicing | 24/2423/24 on a repeat run | 24/2424/24 on a repeat run | ₹31,357vs ₹3.9 lakh | 13×5.8× vs Gemini 3.8 Flash |
| RM messagesRisk & compliance monitoring | 24/2424/24 on a repeat run | 24/2424/24 on a repeat run | ₹18,098vs ₹2.1 lakh | 12× |
| Meeting briefsWealth Assistant | 12/1212/12 on a repeat run | 12/1212/12 on a repeat run | ₹71,646vs ₹5.5 lakh | 7.7×3.1× vs Gemini 3.8 Flash |
| Research monitorResearch Intelligence Monitor | 16/1616/16 on a repeat run | 16/1616/16 on a repeat run | ₹16,431vs ₹2.3 lakh | 14×5.9× vs Gemini 3.8 Flash |
All models we tested
Client instructions
360 ONE model · by Dvar24/24 · ₹0.031
Gemini 3.8 Flash24/24 · ₹0.182
GPT-6.1 Sol24/24 · ₹0.394
Claude Sonnet 5.524/24 · ₹0.679
RM messages
360 ONE model · by Dvar24/24 · ₹0.018
GPT-6.1 Sol24/24 · ₹0.209
Gemini 3.8 Flash23/24 · ₹0.116
Claude Sonnet 5.523/24 · ₹0.345
Meeting briefs
360 ONE model · by Dvar12/12 · ₹0.072
Gemini 3.8 Flash12/12 · ₹0.226
GPT-6.1 Sol12/12 · ₹0.554
Claude Sonnet 5.511/12 · ₹0.985
Research monitor
360 ONE model · by Dvar16/16 · ₹0.016
Gemini 3.8 Flash16/16 · ₹0.097
GPT-6.1 Sol16/16 · ₹0.233
Claude Sonnet 5.515/16 · ₹0.398
76 synthetic cases, checked in code: directional, not a benchmark. 360 ONE's models are open models (GLM, DeepSeek, Gemma) chosen per workflow by Dvar Evals and given each desk's rules; each model shown at the better of two runs, plus a repeat run. Costs at list prices: Dvar's rupee list, OpenRouter for frontier. Demo runs used public endpoints; in production, models run on Dvar's India servers or yours.