Arcee AILanguage model
Trinity Mini
Trinity Mini has no overall index value yet because there are too few independent benchmark results.
28out of 100same as last month
based on search, usage and attention
Summary
Trinity Mini combines 26 billion total parameters with about 3 billion active, a 128K context, and reasoning post-training. Arcee trained the medium variant on roughly 10 trillion tokens with math and coding data. The weights use OpenMDW-1.1, with Transformers and vLLM documented as serving runtimes.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 8 published benchmark results
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
SimpleQA (Show measurement, test conditions, and source)
- Source value
- 8.9
- Score
- 8.9
- Metric
- score as shown
- Unit
- No unit provided
- Benchmark version
- SimpleQA; version/settings not stated
- Category
- general
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Arcee Trinity Mini official model-card benchmark chart
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact Trinity Mini chart values; chart does not specify prompt, sampling settings, run dates or score units.
- Context
- Scores were read from the publisher's static chart and are not independent reruns; the chart omits a full protocol.
MuSR (Show measurement, test conditions, and source)
- Source value
- 63.49
- Score
- 63.49
- Metric
- score as shown
- Unit
- No unit provided
- Benchmark version
- MuSR; version/settings not stated
- Category
- general
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Arcee Trinity Mini official model-card benchmark chart
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact Trinity Mini chart values; chart does not specify prompt, sampling settings, run dates or score units.
- Context
- Scores were read from the publisher's static chart and are not independent reruns; the chart omits a full protocol.
MMLU zero-shot (Show measurement, test conditions, and source)
- Source value
- 84.95
- Score
- 84.95
- Metric
- zero-shot score as shown
- Unit
- No unit provided
- Benchmark version
- MMLU; zero-shot; snapshot not stated
- Category
- general
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Arcee Trinity Mini official model-card benchmark chart
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact Trinity Mini chart values; chart does not specify prompt, sampling settings, run dates or score units.
- Context
- Scores were read from the publisher's static chart and are not independent reruns; the chart omits a full protocol.
Math-500 (Show measurement, test conditions, and source)
- Source value
- 92.1
- Score
- 92.1
- Metric
- score as shown
- Unit
- No unit provided
- Benchmark version
- Math-500; settings not stated
- Category
- general
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Arcee Trinity Mini official model-card benchmark chart
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact Trinity Mini chart values; chart does not specify prompt, sampling settings, run dates or score units.
- Context
- Scores were read from the publisher's static chart and are not independent reruns; the chart omits a full protocol.
GPQA-Diamond (Show measurement, test conditions, and source)
- Source value
- 58.55
- Score
- 58.55
- Metric
- score as shown
- Unit
- No unit provided
- Benchmark version
- GPQA-Diamond; settings not stated
- Category
- general
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Arcee Trinity Mini official model-card benchmark chart
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact Trinity Mini chart values; chart does not specify prompt, sampling settings, run dates or score units.
- Context
- Scores were read from the publisher's static chart and are not independent reruns; the chart omits a full protocol.
BFCL v3 (Show measurement, test conditions, and source)
- Source value
- 59.67
- Score
- 59.67
- Metric
- score as shown
- Unit
- No unit provided
- Benchmark version
- BFCL v3; metric details not stated
- Category
- general
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Arcee Trinity Mini official model-card benchmark chart
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Vendor-reported result. Exact Trinity Mini chart values; chart does not specify prompt, sampling settings, run dates or score units.
- Context
- Scores were read from the publisher's static chart and are not independent reruns; the chart omits a full protocol.
n8n Official AI Benchmark (Show measurement, test conditions, and source)
- Source value
- 56
- Score
- 56
- Metric
- overall publisher composite score
- Unit
- points
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- Official n8n AI Benchmark
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Official benchmark leaderboard. Official n8n AI Benchmark. The publisher says it ranks LLMs by how they work in n8n; the Trinity Mini page states that scores have a maximum of 100 points. The opened sources do not state the workflow mix, aggregation formula, provider/model snapshot, row evaluation date, or sample count.
- Context
- This is an n8n-published workflow/product-system composite, not an independent isolated-model evaluation. Its aggregation is not reconstructed or rescaled.
MK-NIAH (@128K) (Show measurement, test conditions, and source)
- Source value
- 0.888
- Score
- 0.888
- Metric
- MK-NIAH score at 128K context
- Unit
- No unit provided
- Category
- long-context retrieval
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Arcee Trinity Large Technical Report (arXiv:2602.17004v1)
- Status
- active
- Retrieved at
- 2026-10-20
- Methodology
- Vendor-reported result. The report identifies Trinity Mini (26B total parameters, 3B active) and says its final iteration was trained at sequence length 128K for a 128K target context. It does not state an immutable model commit, benchmark sample count, or detailed scoring protocol.
- Context
- The published model label matches the assigned Trinity Mini alias, but the report does not identify the immutable released checkpoint; retain this revision-binding caveat for review.
Popularity compared
How often the model is searched for, used and mentioned, compared with every model of its category.
Popularity index
The 100 most popular models of this category, 0 to 100, higher is more popular
20 of 101 models
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 7, 2026
Trinity MiniLlama 3.1 8B 28Phi-3 Medium (14B) 28Qwen 3 235B A22B Thinking 2507 28Trinity Large Thinking 28
Method changes in the timeline, a jump at these points is no market event:
- 1November 1, 2026: Value only with search or usage data, 30-day attention
What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.
Similar models
More models of the same kind as Trinity Mini, chosen by model line, developer, size, price and release date.
- Virtuoso LargeArcee AI · same developer
- Trinity Nano PreviewArcee AI · same developer
- Trinity Large PreviewArcee AI · same developer
- Trinity Large ThinkingArcee AI · same developer
- Qwen 3.5 27BAlibaba · similar size
- Gemma 4 26B A4BGoogle · similar size
- Qwen 3.6 27BAlibaba · similar size
- Gemma-SEA-LION-v4-27B-ITAI Singapore · similar size
All specifications
Architecture
Input and output
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
Frequently asked questions about Trinity Mini
How large is the context window of Trinity Mini?
131,072 tokens. That is how much text Trinity Mini reads at once, about 200 A4 pages.
Page count as a rule of thumb from Artificial Analysis
Can I run Trinity Mini locally?
Yes, the weights are open.
Stated by Arcee AI
- Open weights
- yes
- License
- OpenMDW-1.1
- Parameters
- 26B
- active per token
- 3B
Sources and data date
Show all 12 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.
Without vendor claims and historical or disputed results; those stay labelled in the single values.
| Type | Evidence and data date |
|---|---|
| Input | Documented modalities. Retrieved October 27, 2026. |
| Output | Documented modalities. Retrieved October 27, 2026. |
| Research status | Research date October 15, 2026. 4 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Not published | evaluation-details Note The chart does not identify all prompting, sampling, dates or score units. These values remain publisher-reported only. |
| Additional source | Hugging Face · Arcee AI Trinity Mini model card · Editorial description reviewed October 2, 2026 |
| Additional source | Arcee Trinity Mini official model-card benchmark chart (retrieved October 15, 2026) |
| Additional source | Official n8n AI Benchmark (retrieved October 19, 2026) (unavailable on November 2, 2026) https://n8n.io/ai-benchmark/trinity-mini/ |
| Additional source | Arcee Trinity Large Technical Report (arXiv:2602.17004v1) (retrieved October 20, 2026) |
| Additional source | Model metadata (retrieved October 15, 2026) |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Ollama: cumulative downloads for the popularity index |
| Additional source | Replicate: cumulative runs for the popularity index |
| Additional source | Google Trending Now: limited feed selection for Germany and the US |
| Additional source | Creative Commons Attribution 4.0 |