#58 Step 3.5 Flash
medium- Cost
- $0.008
- Time
- 277.1s
- Tokens
- 23,695 tok
Stepfun
Release: 2026-02-01
Tested on: 2026-05-22 00:30
stepfun/step-3.5-flash::medium
196B total (11B active)MoEOpen source
(medium)
(none)
Summary
Step 3.5 Flash scores 7.4 on AI BENCHY and ranks #58. It has 9.3 reliability, a 65.8% pass rate, $0.015 total cost, and 43.29s average response time.
What makes Step 3.5 Flash unique: It stands out most in Combined, where it ranks #1, while Coding is its weakest area at #18. Its total benchmark cost is unusually low for its score range.
Model facts
Researched on 2026-08-12
- Parameters
- 196B total (11B active)
- Architecture
- MoE
- Availability
- Open source
- License
- Apache-2.0
7.4
Consistency
8.7
9.3
$0.015
Total Output Tokens
263,806
Total Input Tokens
0
Input Price
$0.090 / 1M
Output Price
$0.300 / 1M
Flaky tests
3
Flaky tests had mixed outcomes across runs (at least one pass and one fail).
Hamster playing table tennis
Prompt: Create a detailed SVG illustration of a hamster playing table tennis.
Run history
| Tested on | Score | Reliability | Tests Correct | Total Cost | Compare |
|---|---|---|---|---|---|
| 2026-08-14 02:50 Re-test | 5.9 | 10.0 | $0.132 ↑ | Compare | |
| 2026-07-16 23:25 New test added | 6.0 | 9.2 | $0.108 ↑ | Compare | |
| 2026-06-04 14:12 New test added | 7.2 | 8.9 | $0.070 ↑ | Compare | |
| 2026-05-22 00:30 Suite changed | 7.4 | 9.3 | $0.015 | Current run | |
| 2026-05-08 15:30 Suite changed | 7.6 | 10.0 | $0.011 | Compare | |
| 2026-04-11 00:35 First recorded run | 7.9 | N/A | $0.000 | Compare |
This run used a different benchmark suite. Keep suite changes in mind when reading historical movement.
Run comparison
| Run | Benchmark coverage | Score | Consistency | Reliability | Tests Correct | Flaky tests | Total Output Tokens | Total Input Tokens | Total Cost | Response Time (avg) |
|---|---|---|---|---|---|---|---|---|---|---|
| 2026-05-22 00:30 · Suite changed | 57/57 attempts | 7.4 | 8.7 | 9.3 | 11/19 | 3 | 263,806 | 0 | $0.015 | 43.29s |
| 2026-04-11 00:35 · First recorded run | 49/51 attempts | 7.9 | 9.1 | N/A | 11/17 | 2 | 227,511 | 0 | $0.000 | 26.78s |
| Difference | — | -0.5 | -0.4 | 0 | +1 | +36295 | 0 | +$0.015 | +16508ms |
Benchmark coverage differs: 57/57 attempts (Target: 3 repeats per test) versus 49/51 attempts (Target: 3 repeats per test). Totals and repeat-sensitive metrics are not directly comparable.
These two runs used different benchmark suites, so the deltas reflect both model changes and suite changes.
Charts
Choose the first model, then click a second model to open a side-by-side page.
Top Models by Score
Score vs Total Cost
Response Time (avg)
Score vs Response Time (avg)
Total Output Tokens
Score vs Total Output Tokens
Quick Compare
Step 3.5 FlashmediumvsQwen3.5-35B-A3BnoneStep 3.5 FlashmediumvsDots 3 Note PreviewmediumFree AvailableStep 3.5 FlashmediumvsQwen3.5-FlashnoneStep 3.5 FlashmediumvsGLM 5.3 FlashlowStep 3.5 FlashmediumvsGPT-5 NanomediumStep 3.5 FlashmediumvsGPT-5.4 MininoneStep 3.5 FlashmediumvsGPT-5.6 TerranoneStep 3.5 FlashmediumvsSeed 2.1 TurbononeStep 3.5 FlashmediumvsGemini 2.5 FlashnoneStep 3.5 FlashmediumvsGPT-5.4none
Category Breakdown
| Category | Score | Consistency | Tests Correct |
|---|---|---|---|
| Anti-AI Tricks | 10.0 | 10.0 | |
| Coding | 3.0 | 6.0 | |
| Combined | 10.0 | 10.0 | |
| Data parsing and extraction | 10.0 | 10.0 | |
| Domain specific | 5.3 | 7.2 | |
| General Intelligence | 5.5 | 10.0 | |
| Instructions following | 8.5 | 6.8 | |
| Puzzle Solving | 5.3 | 10.0 | |
| Tool Calling | 10.0 | 10.0 | |
| Trivia | 3.0 | 10.0 |