Summary
Bison 1 scores 2.2 on AI BENCHY and ranks #526. It has 7.5 reliability, a 23.2% pass rate, $0.004 total cost, and 680ms average response time.
What makes Bison 1 unique: It stands out most in Puzzle Solving, where it ranks #1, while Anti-AI Tricks is its weakest area at #14.
Model facts
Researched on 2026-11-22
- Parameters
- 31B
- Architecture
- Dense
- Availability
- Open source
- License
- MIT AND Apache-2.0
Base Labs decision model fine-tuned from dense Gemma 4 31B. Weights are public: base under Apache-2.0, Baseten changes under MIT. Served as a typed decision endpoint with no reasoning controls; training data (6.71M decision examples) is not released. Released 2026-11-20.
2.2
Consistency
4.1
7.5
$0.004
Total Output Tokens
36
Total Input Tokens
37,001
Input Price
$0.100 / 1M
Output Price
$0.000 / 1M
Cache Read Price
N/A
Cache Write Price
N/A
Cache prices apply to input tokens. Reads reuse cached prompts; writes store them and can cost extra. Output tokens use the output price.
Wrong Tests: 8
Attempt pass rate: 23.2%
Benchmark coverage: 36/69 attempts. Choices supplied by adapters. Supports 12/23 tests. Other tests are unsupported, not failed. Score uses the full suite.
Flaky tests
3
Flaky tests had mixed outcomes across runs (at least one pass and one fail).
Charts
Choose the first model, then click a second model to open a side-by-side page.
Score vs Total Cost
Response Time (avg)
Score vs Response Time (avg)
Total Output Tokens
Score vs Total Output Tokens
Quick Compare
Category Breakdown
| Category | Score | Consistency | Tests Correct |
|---|---|---|---|
| Agentic | 0.0 | 0.0 | |
| Anti-AI Tricks | 3.9 | 5.4 | |
| Coding | 2.9 | 3.7 | |
| Combined | 0.0 | 0.0 | |
| Data parsing and extraction | 6.4 | 5.9 | |
| Domain specific | 5.3 | 10.0 | |
| General Intelligence | 0.0 | 0.0 | |
| Instructions following | 1.5 | 5.0 | |
| Puzzle Solving | 3.3 | 3.3 | |
| Tool Calling | 0.0 | 0.0 | |
| Trivia | 0.0 | 0.0 |