Skip to content

Trinity Large Preview

View Status

Trinity Large Preview is a 400B-parameter sparse mixture-of-experts model from Arcee AI that activates 13B parameters per forward pass, targeting math, coding, and multi-step agent workloads across a context window of 131K tokens.

Tool Use
index.ts
import { streamText } from 'ai'


const result = streamText({
  model: 'arcee-ai/trinity-large-preview',
  prompt: 'Why is the sky blue?'
})

What To Consider When Choosing a Provider

  • Configuration: Stream long analytical outputs to improve time-to-first-token. At $0.25 per million input tokens and $1 per million output tokens, compare spend to your latency budget before you scale traffic.
  • Zero Data Retention: AI Gateway does not currently support Zero Data Retention for this model. See the documentation for models that support ZDR.
  • Authentication: AI Gateway authenticates requests using an API key or OIDC token. You do not need to manage provider credentials directly.

When to Use Trinity Large Preview

Best For

  • Multi-step agent pipelines: The model plans, calls tools, and synthesizes results over many turns
  • Sustained math and coding: Tasks that need continuous reasoning across a long context
  • Large-scale code work: Generation or debugging where the model must follow logic across large files or refactors
  • Long-context analysis: Ingesting a large corpus and producing structured conclusions

Consider Alternatives When

  • Short single-turn requests: A smaller or faster model may match quality at lower cost
  • Generally available contract: Preview terms don't fit teams needing a fixed long-term API contract
  • Latency-dominant tasks: Simpler models suffice when deep multi-step reasoning isn't required

Conclusion

Trinity Large Preview brings Arcee AI's large MoE stack to AI Gateway for agentic, math, and coding workloads. If you need long-context reasoning and can accept preview terms, run it through your own benchmarks on AI Gateway.

Frequently Asked Questions

  • What kinds of tasks is Trinity Large Preview explicitly designed for?

    Math, coding, and complex multi-step agent workflows. The release notes also emphasize efficient extended multi-turn use with high inference throughput.

  • Why is this model labeled a "preview" release?

    It ships before general availability. Expect changes while Arcee AI finalizes production behavior, pricing, and versioning.

  • How does Trinity Large Preview differ from Trinity Mini?

    Trinity Large Preview is a 400B-parameter MoE with 13B active per forward pass, targeting deep reasoning on math, coding, and multi-step agent tasks. Trinity Mini is a 26B-parameter MoE with 3B active parameters, tuned for lean inference and volume. Pick Mini when cost per token is the binding constraint; pick this model when you need the larger parameter space.

  • Do I need an Arcee AI account to use this model through AI Gateway?

    No. Use your AI Gateway API key or an OIDC token. You don't need a separate provider account.

  • Can I use this model with the AI SDK?

    Yes. Set model to arcee-ai/trinity-large-preview in the AI SDK's streamText or generateText call. AI Gateway also exposes OpenAI Chat Completions, OpenAI Responses, Anthropic Messages, and OpenResponses-compatible interfaces.

  • Does AI Gateway provide observability for requests to Trinity Large Preview?

    Yes. Token usage, latency, and cost show in your AI Gateway dashboard for each request without extra instrumentation.