Search Capability Leaderboard
DeepSeek V4.1 Flash (1017)
DeepSeekOpen weights
#10 in Search Intelligence · #4 in Search Efficiency
Search Intelligence Score
68.9
+36.0 lift from search
050100
- Lift from search
- +36.0
- 32.9 without search
- Cost per 1K tasks
- $125
- #4 in Search Efficiency
- Time per task
- 540s
- Search plus inference time to complete one task
Accuracy by benchmark
Accuracy on each eval, with search (solid) and without (light).
- DSQA84.6 vs 46.6
- HLE53.0 vs 40.0
- WISER69.0 vs 12.0
Cost per 1K tasks
How the cost splits between model inference, search calls, and page extraction.
- Inference
- $58.2
- 46%
- Search
- $21.6
- 17%
- Extract
- $45.4
- 36%
Compared with
Other models from the same provider alongside models that score closest.
- DS Flash 4.1 101768.9Cost $125 · Lift +36.0 · 540s
- DS Flash 4.162.7Cost $35.8 · Lift +35.7 · 549s
- DS Pro58.4Cost $110 · Lift +27.9 · 383s
- Pareto 26.1068.6Cost $116 · Lift +33.8 · 436s
- Ember-167.9Cost $335 · Lift +38.0 · 487s
| Model | Score | Lift | Cost per 1K tasks | Time per task |
|---|---|---|---|---|
| DeepSeek V4.1 Flash (1017) | 68.9 | +36.0 | $125 | 540s |
| DS Flash 4.1DeepSeek V4.1 Flash | 62.7 | +35.7 | $35.8 | 549s |
| DS ProDeepSeek V4 Pro | 58.4 | +27.9 | $110 | 383s |
| Pareto 26.10Pareto 26.10 | 68.6 | +33.8 | $116 | 436s |
| Ember-1Ember-1 | 67.9 | +38.0 | $335 | 487s |
Search Intelligence Score averages accuracy on DSQA, HLE, and WISER with equal weight; cost and time are averaged the same way, with cost shown per 1,000 tasks. Latest update October 18, 2026. Full methodology.
# DeepSeek V4.1 Flash (1017) · Search Capability Leaderboard
- Search Intelligence Score: 68.9 (#10 in Search Intelligence)
- Without search: 32.9 · Lift from search +36.0
- Cost per 1K tasks: $125 (#4 in Search Efficiency)
- Time per task: 540s
- Lab: DeepSeek · Weights: Open weights · Model id: deepseek/deepseek-v4.1-flash-1017
## Accuracy by benchmark
| Suite | With search | Without search |
|---|---|---|
| DSQA | 84.6 | 46.6 |
| HLE | 53.0 | 40.0 |
| WISER | 69.0 | 12.0 |
## Compared with
| Model | Score | Lift | Cost per 1K tasks | Time per task |
|---|---|---|---|---|
| DeepSeek V4.1 Flash (1017) | 68.9 | +36.0 | $125 | 540s |
| DeepSeek V4.1 Flash | 62.7 | +35.7 | $35.8 | 549s |
| DeepSeek V4 Pro | 58.4 | +27.9 | $110 | 383s |
| Pareto 26.10 | 68.6 | +33.8 | $116 | 436s |
| Ember-1 | 67.9 | +38.0 | $335 | 487s |
Latest update October 18, 2026. Methodology: /leaderboard#methodology