Trinity Large Thinking
Arcee AI · released Apr 1, 2026
Trinity Large Thinking is a powerful open source reasoning model from the team at Arcee AI. It shows strong performance in PinchBench, agentic workloads, and reasoning tasks. Launch video: https://youtu.be/Gc82AXLa0Rg?si=4RLn6WBz33qT--B7...
Specification
- Context window
- 262K
- Max output
- 262K
- Knowledge cutoff
- Not stated
- Parameters
- 398.6B
- Licence
- other
- Serving providers
- 2
- Moderated
- No
Intelligence
18.6
56th percentile
Coding
25.8
Coding Index
Agentic
—
Agentic Index
Output speed
67 t/s
Median across providers
Latency
744ms
Time to first token
Cost per task
$0.08
Estimated
Benchmarks
Where the score comes from
The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.
Evaluation scores
Percentage correct · higher is better
- τ²-bench (Telecom)90.1%
- GPQA Diamond75.2%
- IFBench56.3%
- AA-LCR (long context)38.3%
- SciCode36.1%
- Terminal-Bench Hard22.7%
- Humanity's Last Exam15.8%
An evaluation missing from this list was not run for this model — it is not a zero.
View as table
| Evaluation | Score |
|---|---|
| τ²-bench (Telecom) | 90.1% |
| GPQA Diamond | 75.2% |
| IFBench | 56.3% |
| AA-LCR (long context) | 38.3% |
| SciCode | 36.1% |
| Terminal-Bench Hard | 22.7% |
| Humanity's Last Exam | 15.8% |
Against its peers
Intelligence Index · this model highlighted, nearest peers in grey
- GPT-5 NanoOpenAI20.1
- GLM 4.5Z.ai19.7
- Kimi K2 0711Moonshot AI19.7
- GPT-4.1OpenAI19.6
- o3 MiniOpenAI19.2
- o1-proOpenAI19.1
- Trinity Large ThinkingArcee AI18.6
- Sonar Reasoning ProPerplexity18.0
Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.
View as table
| Model | Intelligence |
|---|---|
| GPT-5 Nano | 20.1 |
| GLM 4.5 | 19.7 |
| Kimi K2 0711 | 19.7 |
| GPT-4.1 | 19.6 |
| o3 Mini | 19.2 |
| o1-pro | 19.1 |
| Trinity Large Thinking | 18.6 |
| Sonar Reasoning Pro | 18.0 |
Percentile among all indexed models
Pricing
What it costs to run
List prices per million tokens, plus what one representative task works out to.
List price
- Input / 1M tokens
- $0.22
- Output / 1M tokens
- $0.85
- Cached input / 1M
- $0.06
- Blended 3:1
- $0.378
One task, estimated
$0.08
- Input tokens
- 50,000
- Output tokens
- 80,000
- Profile
- Reasoning
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).
Arena
Head-to-head generation quality
Elo from pairwise judgements, broken out by the kind of thing the model was asked to build.
- Overall Elo
- 1,159
- Win rate
- 41.3%
- Strongest at
- Websites
- Tournaments
- 7,105
Elo by category
Dot position on a 1,050–1,175 scale · Elo has no meaningful zero
- 1,159
- 1,137
- 1,126
- 1,120
- 1,080
- 1,062
Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.
View as table
| Category | Elo | Win rate |
|---|---|---|
| Websites | 1159 | 41.3% |
| 3D scenes | 1137 | 41.3% |
| Data visualisation | 1126 | 39.3% |
| Game development | 1120 | 38.4% |
| UI components | 1080 | 32.6% |
| SVG | 1062 | 35.2% |