IntroducingEngine
Fortytwo Engine is an inference API that runs open-weight models through Swarm Inference.
The same models, more accurate and efficient, at the same per-token price.
Qwen3.8-27B
More coming soonMore models coming soon
Same Qwen.
Better answers.
7% more
accurateon average
vs. base modelUp to 25%
fewer tokensper task
Private
by default
with zero
data retention
Token efficiency
Usage across 450 DeepSWE benchmark runs.
Metric
Base model
Engine
Change
Average tokens per run
Base model
Engine
Change
Base model: 16.3M
Engine: 12.3M
Change: -24.7%
Answer accuracy
Qwen3.8-27B through Fortytwo Engine versus the base model, evaluated under the same conditions.
Benchmark
Base model
Engine
Change
Benchmark
DeepSWE
Base model
Engine
Change
Base model: 42.2
Engine: 47.8
Change: +13.3%
GDP.pdf
Base model: 18.5
Engine: 20.6
Change: +11.4%
GPQA-Diamond
Base model: 90.9
Engine: 95.1
Change: +4.6%
Humanity’s Last Exam
Base model: 34.7
Engine: 35.8
Change: +3.2%
LiveCodeBench v6
Base model: 80.2
Engine: 84.1
Change: +4.9%
SciCode
Base model: 46.8
Engine: 49.0
Change: +4.7%
Model specs
Pricing
- Input
- $0.425 / MTok
- Output
- $2.55 / MTok
- Cache read
- $0.085 / MTok
Context
- Window
- 262,144
- Max output tokens
- 235,929
General
- Reasoning
- Supported
- Input
- Text, Image
- Output
- Text
Rate limits
- Core
- 20 RPM / 500 RPH
- Pro
- 200 RPM / 7,000 RPH
- Scale
- 600 RPM / 20,000 RPH
Endpoints
- v1/chat/completions
- Supported
Supported features
- Function calling
- Supported
- Structured outputs
- Supported
- Image input
- Supported
- Streaming
- Supported
Start your Engine
on Fortytwo Platform
Sign in
