Arcee AILanguage model
Trinity Nano Preview
Trinity Nano Preview has no overall index value yet because there are too few independent benchmark results.
25out of 100-3 vs. last month
based on search, usage and attention
Summary
Trinity Nano Preview brings the Trinity design down to 6 billion total parameters and about 1 billion active. Arcee documents a 128K context, chat tuning, and only about 800 million active non-embedding parameters per token. The model card warns about instability and explicitly says the Preview should be downloaded and run locally rather than hosted.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 2 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.
Medmarks weighted peer-comparison win rate (Show measurement, test conditions, and source)
- Source value
- 34.2
- Score
- 34.2
- Metric
- weighted share of pairwise comparison points across datasets and models; not answer accuracy
- Unit
- %
- Benchmark version
- Medmarks live public leaderboard; row shows Dec 1, 2025, but no run snapshot or configuration is linked
- Direction
- Higher is better
- Source type
- official-leaderboard
- Evaluator
- Medmarks live leaderboard and scoring-methodology note
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Official benchmark leaderboard. Exact public row named Trinity Nano Preview (Arcee AI, TINY), displayed with date Dec 1, 2025. No underlying endpoint ID, weight revision, subset filter, or per-run configuration is exposed.
- Context
- The aggregate is not an accuracy score. The live leaderboard does not link the exact model revision or underlying per-model run, so retain the displayed exact-name row without treating it as a pinned checkpoint evaluation.
Medmarks-V weighted mean win rate, paper v1 (Show measurement, test conditions, and source)
- Source value
- 0.3596
- Score
- 0.3596
- Metric
- Dataset-weighted mean pairwise win rate against peer models; not answer accuracy
- Unit
- ratio
- Benchmark version
- Medmarks paper v1 Table 2, Medmarks-V
- Category
- medical-llm-evaluation
- Direction
- Higher is better
- Source type
- independent-evaluator
- Evaluator
- Medmarks independent evaluation paper v1
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact paper row Trinity Nano Preview, size Tiny; Table 3 configuration identifies Trinity Nano 6B-A1B with temperature 0.5, top_p 0.95, top_k 50. The score is the Medmarks-V dataset-weighted mean of pairwise peer-comparison wins; no immutable checkpoint revision is given.
- Context
- This historical paper-v1 Medmarks-V result is distinct from the existing 34.2% public leaderboard observation. The live page exposes separate Combined, Verified, and Open-Ended views and does not link a run; no difference is interpreted as a trend or contradiction. The paper supplies no model-specific sample count or confidence interval. Its separate Num Regr. statistic of 107 is not a quality-score denominator. The paper does not pin the checkpoint revision.
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 Nano PreviewFalcon 180B 25Index-Homura-9B 25ERNIE 4.5 VL 424B A47B 25Index-Homura-2B 25
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 Nano Preview, chosen by model line, developer, size, price and release date.
- Trinity MiniArcee AI · same developer
- Trinity Large PreviewArcee AI · same developer
- Trinity Large ThinkingArcee AI · same developer
- Virtuoso LargeArcee AI · same developer
- Gemma 4 E2B ITGoogle · similar size
- Gemma 3n E2B ITGoogle · similar size
- Qwen 3 VL 8BAlibaba · similar size
- Qwen3.5 4BAlibaba · 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 Nano Preview
How large is the context window of Trinity Nano Preview?
131,072 tokens. That is how much text Trinity Nano Preview reads at once, about 200 A4 pages.
Page count as a rule of thumb from Artificial Analysis
Can I run Trinity Nano Preview locally?
Yes, the weights are open.
Stated by Arcee AI
- Open weights
- yes
- License
- OpenMDW-1.1
- Parameters
- 6B
- active per token
- 1B
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. 5 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Unresolved | medmarks-paper-checkpoint-revision-and-model-specific-n Note The paper's exact-name row and generation configuration support the profile association, but it does not record the immutable Arcee model-card revision, model-specific number of scored examples, or a confidence interval. No number is inferred from the separate regression-token statistic. |
| Additional source | Hugging Face (retrieved October 15, 2026; October 16, 2026) · Arcee AI Trinity Nano Preview model card · Editorial description reviewed October 2, 2026 |
| Additional source | Medmarks live leaderboard and scoring-methodology note (retrieved October 15, 2026; October 16, 2026) |
| Additional source | Medmarks independent evaluation paper v1 (retrieved October 16, 2026) |
| Additional source | Model metadata (retrieved October 15, 2026) |
| Additional source | Model metadata (retrieved October 16, 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 |