Approved models
17
Records published from this collection
Provider information
lab
Review approach
Verify each model’s source, terms, capabilities, and current availability
DeepSeek publishes downloadable model weights and a hosted API. Its current API identifiers and service notices matter when maintaining an application that was built around an earlier release.
DeepSeek V4.1 Flash adds native image understanding. The current Flash identifier is deepseek-flash; older V4 Flash names can resolve through compatibility routes. Record the requested identifier and the model behind it when comparing a migration.
The architecture has different active-parameter counts for input encoding and output decoding. Do not use one active count to estimate the complete weight download or server memory requirement.
The V4.1 Flash checkpoint uses MIT licensing. A local deployment requires a compatible runtime and suitable hardware; a managed route has its own quotas and commercial terms. Keep the exact route with the test record.
For visual work, include small text, charts and contradictory image/prompt pairs. For coding, verify the patch with executable checks and inspect unrelated changes. Measure retries, output validation failures and time spent correcting the result.
DeepSeek documents peak and off-peak pricing. A single stored price cannot describe that schedule. Earlier announcements can also be superseded by newer service information. Check the current API documentation before changing an endpoint, budgeting a workload or deciding that an older route has disappeared.
Review each record individually; collection membership does not imply a recommendation or guarantee of availability.
| DeepSeek V4.1 Flash DeepSeek | 24.8 | 1.05M | $0.15 | $0.60 | — |
| DeepSeek V4 Flash Vision Experimental DeepSeek | 56.9 | 1.05M | — | — | — |
| DeepSeek V4 Pro 0813 DeepSeek | — | 1M | — | — | — |
| DeepSeek V4 Flash 0731 DeepSeek | 71.6 | 1M | $0.1596 | $0.399 | — |
| DeepSeek V4 Flash DeepSeek | 71.1 | 1M | $0.14 | $0.28 | — |
| DeepSeek V4 Pro DeepSeek | 71.1 | 1M | $0.435 | $0.87 | — |
| DeepSeek V4 Pro 0423 DeepSeek | — | 1M | — | — | — |
| DeepSeek Chat DeepSeek | 50.3 | 1M | $0.14 | $0.28 | — |
| DeepSeek Reasoner DeepSeek | 65.3 | 1M | $0.14 | $0.28 | — |
| DeepSeek V3.2 DeepSeek | — | 128K | — | — | — |
| DeepSeek-V3.1 DeepSeek | — | 131.07K | — | — | — |
| DeepSeek R1 0528 Qwen3 8B DeepSeek | — | 131.07K | — | — | — |
| DeepSeek R1 0528 DeepSeek | — | 163.84K | — | — | — |
| DeepSeek V3 0324 DeepSeek | — | 163.84K | — | — | — |
| DeepSeek-R1 DeepSeek | 49.7 | 128K | $0.70 | $2.50 | — |
| DeepSeek-R1-Distill-Qwen-32B DeepSeek | — | 131.07K | — | — | — |
| DeepSeek-V3 DeepSeek | — | 131.07K | — | — | — |
1–17 of 17 models · select a column to sort
About this order: Recommended shows our editorial shortlist first, followed by recent releases. The spec score is a 0–100 figure SyncDev calculates from each vendor's published specifications — context window, reasoning support, input modalities, tool calling, maximum output and how recently the model shipped. It measures breadth of capability, not benchmark performance, so a higher-scoring model is not automatically the better choice for your task.