Skip to main content

Model profile

Trinity-Large-Thinking

Arcee AICurrentReleased Mar 10, 2026
Data verified
Overall Score
52.32Public #100 of 200Verified #55 of 99
Arena Elo
1369
Eligible category ranks
1of 8
Price (1M tokens)
$0.25 in / $0.9 out
API pricing
Speed
Not listed
Context
512K

Evidence coverage

19 of 323 tracked benchmarks are published. 1 is verified and 18 provisional. 7 of 8 categories are measured.

Updated July 23, 2026Methodology
Published / tracked
19 / 323
Verified
1
Provisional
18
Categories with evidence
7 / 8

Evidence by category

  • Agentic4 benchmarks
    Mixed evidence
  • Coding2 benchmarks
    Reported
  • Reasoning2 benchmarks
    Reported
  • Knowledge8 benchmarks
    Reported
  • Math1 benchmark
    Reported
  • Multilingual0 benchmarks
    Not measured
  • Multimodal1 benchmark
    Reported
  • Inst. Following1 benchmark
    Reported
Open WeightSelf-hostReasoning
Confidence:
Low
thinking

Trinity-Large-Thinking ranks #100 out of 200 models on the public leaderboard with an overall score of 52.32/100. It also ranks #55 out of 99 on the verified leaderboard. While not a frontier model, it offers specific advantages depending on the use case.

Trinity-Large-Thinking is a open weight model with a 512K token context window. It uses explicit chain-of-thought reasoning, which typically improves performance on math and complex reasoning tasks at the cost of higher latency and token usage.

Trinity-Large-Thinking sits inside the Trinity Large family alongside Trinity-Large-Preview. This profile currently has 19 of 323 tracked benchmarks. BenchLM only exposes non-generated benchmark rows publicly, so missing categories stay blank until a sourced evaluation is available.

Its strongest category is Coding (#87). This performance profile makes it particularly well-suited for software development and code generation tasks.

Peer position

Exact provisional scores and ranks for the closest listed peers. A score can appear before a model clears the evidence threshold for a rank, so equal scores can have different rank states.

Range 51.4752.94

  1. Nemotron 3 Nano 30B
    NVIDIA
    #9852.94
    Nemotron 3 Nano 30B is #98 with a score of 52.94.
    Compare
  2. GPT-5.1-Codex
    OpenAI
    #9952.72
    GPT-5.1-Codex is #99 with a score of 52.72.
    Compare
  3. Trinity-Large-ThinkingCurrent model
    Arcee AI
    #10052.32
    Trinity-Large-Thinking is #100 with a score of 52.32.
  4. Qwen2.5-72B
    Alibaba
    #10152.15
    Qwen2.5-72B is #101 with a score of 52.15.
    Compare
  5. Llama 3.1 405B
    Meta
    #10251.71
    Llama 3.1 405B is #102 with a score of 51.71.
    Compare
  6. DeepSeek-R1
    DeepSeek
    #10351.67
    DeepSeek-R1 is #103 with a score of 51.67.
    Compare
  7. Qwen3.6-35B-A3B
    Alibaba
    #10451.47
    Qwen3.6-35B-A3B is #104 with a score of 51.47.
    Compare

Category percentile

More

Relative position among models eligible for each sourced category. A higher percentile means a stronger position within that category's ranked cohort; 100 is highest.

  1. Coding29%
    Eligible cohort rank #87 of 122Category score 46.6

Category evidence

Scores and ranks appear only where this model has published benchmark evidence. Categories without displayable source records remain not measured.

Category scores, ranks, weighting, benchmark coverage, and evidence status
CategoryScore
AgenticWeight 22%4 benchmarksMixed sourcesScore pending
CodingRank #87 of 122Percentile 29thWeight 20%2 benchmarksReported46.6
ReasoningWeight 17%2 benchmarksReportedScore pending
KnowledgeWeight 12%8 benchmarksReportedScore pending
MathWeight 5%1 benchmarkReportedScore pending
MultilingualWeight 7%0 benchmarksNot measuredNot measured
MultimodalWeight 12%1 benchmarkReportedScore pending
Inst. FollowingRank Not rankedWeight 5%1 benchmarkReported52.3

Chatbot Arena performance

Scroll horizontally to inspect confidence intervals and vote counts.

Chatbot Arena Elo, confidence interval, and vote count by evaluation view
ViewEloConfidence intervalVotes
Text Overall1369±4.729,139
Coding1414±7.68,159
Math1384±15.21,618
Instruction Following1358±7.09,786
Creative Writing1336±9.64,548
Multi-turn1349±9.15,385
Hard Prompts1386±5.719,102
Hard Prompts (English)1396±7.19,515
Longer Query1367±6.912,281

Benchmark Details

Rows below have a displayable published verification record. Each source link and provenance note remains in the page HTML while its category is closed. Source-unverified manual rows and generated rows stay hidden.

Agentic4 benchmarks
τ²-bench resultsReported

τ²-Bench Tool-Agent-User Evaluation

90.1%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Gert LabsBenchmark exact

Gert Labs Composite Game Benchmark

32.55%Display only
Source: Gert Labs rankingsProvenance: Gert Labs reports this composite leaderboard score in the public rankings API. BenchLM scales the source gscore from 0-1 to 0-100 and stores it as a display-only agentic benchmark.
GDPval-AAReported

GDPval-AA normalized

2.7%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
GDPval-AAReported
554Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Coding2 benchmarks
SWE-bench Verified*Secondary exact

SWE-bench Verified (mini-swe-agent-v2)

63.2%Display only
Source: Arcee Trinity-Large-Thinking comparison tableProvenance: Secondary exact
AA-SciCodeReported

Artificial Analysis SciCode

36.1%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Reasoning2 benchmarks
AA-LCRReported

Artificial Analysis Long Context Reasoning

33.0%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
CritPtReported

Critical Physics Tasks

0.9%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Knowledge8 benchmarks
GPQA-DSecondary exact

GPQA Diamond

76.3%Display only
Source: Arcee Trinity-Large-Thinking comparison tableProvenance: Secondary exact
MMLU-Pro (Arcee)Secondary exact

MMLU-Pro first-party comparison snapshot

83.4%Display only
Source: Arcee Trinity-Large-Thinking comparison tableProvenance: Secondary exact
Artificial Analysis Intelligence IndexReported
24.5%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-GPQA DiamondReported

Artificial Analysis GPQA Diamond

75.2%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-HLEReported

Artificial Analysis Humanity's Last Exam

14.7%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience IndexReported

Artificial Analysis Omniscience Index

-44.2%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience AccuracyReported

Artificial Analysis Omniscience Accuracy

22.8%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
AA-Omniscience Hallucination RateReported

Artificial Analysis Omniscience Hallucination Rate

86.6%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.
Math1 benchmark
AIME25 (Arcee)Secondary exact

AIME25 first-party comparison snapshot

96.3%Display only
Source: Arcee Trinity-Large-Thinking comparison tableProvenance: Secondary exact
Multimodal1 benchmark
Design Arena WebsiteReported

Design Arena Website Elo

1165Display only
Source: OpenRouter model benchmarksProvenance: Display-only Design Arena Website Elo synced from OpenRouter model benchmark metadata. It is excluded from BenchLM weighted scoring.
Inst. Following1 benchmark
AA-IFBenchReported

Artificial Analysis IFBench

56.3%Display only
Source: Artificial Analysis model benchmarksProvenance: Display-only row synced from the current Artificial Analysis model payload. It is excluded from BenchLM weighted scoring.

Trinity Large Family

Thinking

Frequently Asked Questions

How does Trinity-Large-Thinking perform overall in AI benchmarks?

Trinity-Large-Thinking has 19 published benchmark scores on BenchLM, but it does not yet have enough non-generated coverage to receive a global overall rank.

Is Trinity-Large-Thinking good for knowledge and understanding?

Trinity-Large-Thinking has visible benchmark coverage in knowledge and understanding, but BenchLM does not currently assign it a global category rank there.

Is Trinity-Large-Thinking good for coding and programming?

Trinity-Large-Thinking ranks #87 out of 122 models in coding and programming benchmarks with an average score of 46.6. There are stronger options in this category.

Is Trinity-Large-Thinking good for mathematics?

Trinity-Large-Thinking has visible benchmark coverage in mathematics, but BenchLM does not currently assign it a global category rank there.

Is Trinity-Large-Thinking good for reasoning and logic?

Trinity-Large-Thinking has visible benchmark coverage in reasoning and logic, but BenchLM does not currently assign it a global category rank there.

Is Trinity-Large-Thinking good for agentic tool use and computer tasks?

Trinity-Large-Thinking has visible benchmark coverage in agentic tool use and computer tasks, but BenchLM does not currently assign it a global category rank there.

Is Trinity-Large-Thinking good for multimodal and grounded tasks?

Trinity-Large-Thinking has visible benchmark coverage in multimodal and grounded tasks, but BenchLM does not currently assign it a global category rank there.

Is Trinity-Large-Thinking good for instruction following?

Trinity-Large-Thinking has visible benchmark coverage in instruction following, but BenchLM does not currently assign it a global category rank there.

Is Trinity-Large-Thinking open source?

Yes, Trinity-Large-Thinking is an open weight model created by Arcee AI, meaning it can be downloaded and run locally or fine-tuned for specific use cases.

Which sibling models are related to Trinity-Large-Thinking?

Trinity-Large-Thinking belongs to the Trinity Large family. Related variants on BenchLM include Trinity-Large-Preview.

Does Trinity-Large-Thinking have full benchmark coverage on BenchLM?

Not yet. Trinity-Large-Thinking currently has 19 published benchmark scores out of the 323 benchmarks BenchLM tracks. BenchLM only exposes non-generated public benchmark rows, so missing categories stay blank until a sourced evaluation is available.

What is the context window size of Trinity-Large-Thinking?

Trinity-Large-Thinking has a published context window of 512K, which determines how much text it can process in a single interaction.

Last updated: July 23, 2026 · Runtime metrics stay blank until BenchLM has a sourced snapshot.

Choose with this week’s evidence

Join 2,000+ readers for ranking moves, new releases, pricing changes, and the evidence behind them.

Free. One email per week.