Saba vs Step 3.7 Flash
Updated October 21, 2026. 3 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Neither model has a coding score yet, so the two cannot be ranked on coding.
Saba has the edge where output tokens cost 1.9x less, and where what matters is AA Intelligence Index or GPQA Diamond (Artificial Analysis run), where it leads.
- Shared results
- 3
- Saba only
- 1
- Step 3.7 Flash only
- 15
- Like-for-like categories
- 2 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
Step 3.7 Flash
Step 3.7 Flash leads, 19.5 against 6.5.
Same board, both measured
Web design
Design Arena: websites
No call
Only Step 3.7 Flash has a Design Arena: websites result, 1196.
Directional only
Long documents
Listed context window
Step 3.7 Flash
Step 3.7 Flash takes 262K tokens against 33K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Saba
Saba $0.0005, Step 3.7 Flash $0.0008 per turn.
List prices
At a glance
- Coding index
- Saba not benchmarked, Step 3.7 Flash not benchmarked.
- Widest category
- Reasoning, 27.8 points apart on average over 2 shared benchmarks.
- Widest single gap
- GPQA Diamond (Artificial Analysis run): Saba 42.4% against Step 3.7 Flash 80.9%, 38.5 points apart on a 0 to 100 scale.
- Output price
- Saba $0.60 against Step 3.7 Flash $1.15 per million output tokens, 1.9x apart.
- Context window
- Saba takes 33K tokens, Step 3.7 Flash 262K.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | Saba | Bars | Step 3.7 Flash | Ties |
|---|---|---|---|---|
| Intelligence | 0 | 1 | ||
| Reasoning | 0 | 2 | ||
| All shared boards | 0 | 3 |
CodingDirectional onlySix of these boards feed the Vibe Coding Index4 boards, none shared
| Benchmark | Saba | Step 3.7 Flash | Leader |
|---|---|---|---|
| AA Coding Index | Not measured | 39.6 | Only Step 3.7 Flash measured |
| SciCode | Not measured | 43.9% | Only Step 3.7 Flash measured |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 35.6% | Only Step 3.7 Flash measured |
| ALE-Bench | Not measured | 694.1 | Only Step 3.7 Flash measured |
| Neither model has a result on the other 34 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Saba | Step 3.7 Flash | Leader |
|---|---|---|---|
| AA Intelligence Index | 6.5 | 19.5 | Step 3.7 Flash |
| Neither model has a result on the other 1 board here | |||
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
| Benchmark | Saba | Step 3.7 Flash | Leader |
|---|---|---|---|
| τ²-Bench Telecom | Not measured | 98.5% | Only Step 3.7 Flash measured |
| Neither model has a result on the other 38 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Saba | Step 3.7 Flash | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | 42.4% | 80.9% | Step 3.7 Flash |
| Humanity's Last Exam (Artificial Analysis run) | 4.3% | 21.4% | Step 3.7 Flash |
| Neither model has a result on the other 19 boards here | |||
MathNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Saba | Step 3.7 Flash | Leader |
|---|---|---|---|
| Neither model has a result on 13 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
| Benchmark | Saba | Step 3.7 Flash | Leader |
|---|---|---|---|
| MMLU Pro (Artificial Analysis run) | 61.1% | Not measured | Only Saba measured |
| Neither model has a result on the other 16 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
| Benchmark | Saba | Step 3.7 Flash | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1196 | Only Step 3.7 Flash measured |
| Design Arena: UI components | Not measured | 1179 | Only Step 3.7 Flash measured |
| Design Arena: game dev | Not measured | 1158 | Only Step 3.7 Flash measured |
| Design Arena: data viz | Not measured | 1172 | Only Step 3.7 Flash measured |
| Design Arena: 3D | Not measured | 1129 | Only Step 3.7 Flash measured |
| Design Arena: all code categories | Not measured | 1182 | Only Step 3.7 Flash measured |
| Design Arena: SVG | Not measured | 1085 | Only Step 3.7 Flash measured |
| Design Arena: ASCII art | Not measured | 1152 | Only Step 3.7 Flash measured |
| Neither model has a result on the other 1 board here | |||
GeneralDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | Saba | Step 3.7 Flash | Leader |
|---|---|---|---|
| IFBench | Not measured | 67.3% | Only Step 3.7 Flash measured |
| Long context reasoning (AA-LCR) | Not measured | 73.7% | Only Step 3.7 Flash measured |
| Neither model has a result on the other 44 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Saba | Step 3.7 Flash | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
Saba costs less
Repository review
50K input + 3K output tokens
Saba costs less
Agent loop
200K input + 10K output tokens
Saba costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | Mistral Saba | StepFun Step 3.7 Flash |
|---|---|---|
| Maker | Mistral AI | StepFun |
| Weights | No public weights listed | Open weights |
| Released | February 17, 2025 | May 28, 2026 |
| Context window | 33K | 262K |
| Input $/Mtok | $0.20 | $0.20 |
| Output $/Mtok | $0.60 | $1.15 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 0 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Saba better than Step 3.7 Flash for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Saba has the edge where output tokens cost 1.9x less, and where what matters is AA Intelligence Index or GPQA Diamond (Artificial Analysis run), where it leads.
Which is cheaper, Saba or Step 3.7 Flash?
Saba costs $0.20 per million input tokens and $0.60 per million output tokens; Step 3.7 Flash costs $0.20 and $1.15. Cost per solved task is known for neither of them.
Which has the bigger context window, Saba or Step 3.7 Flash?
Step 3.7 Flash, with 262K tokens against 33K.
How many benchmarks is the Saba vs Step 3.7 Flash comparison based on?
3 benchmarks have a result for both models as of October 21, 2026. Saba's coding index rests on 0 independent benchmarks and Step 3.7 Flash's on 0. Where only one model has a result, the page says so instead of naming a winner.