Current Instruments : Text Max, Text Prime & Text Standard
The Grid is a real-time spot market for AI inference. Today, three text-to-text Instruments are available for trading: Text Max, frontier models for deep reasoning, long context, and complex workflows. Text Prime, optimized for quality and long-form coherence at a high intelligence floor, and Text Standard, optimized for low latency and high throughput at a lower cost per Unit. Choosing between them depends on whether your workload prioritizes response quality or response speed - and since all markets use the same OpenAI-compatible Consumption API, you can switch between them by changing a single parameter.
Overview
The initial Text-to-Text Instruments available on The Grid are:
Text Standard: Optimized for speed and throughput.
Very low time to first token.
High streaming tokens per second.
Designed for short to medium outputs.
Intelligence floor that is good enough for many production workloads.
Text Prime: Optimized for quality and long form coherence.
Higher minimum intelligence benchmark score.
Larger maximum output size.
Accepts somewhat slower time to first token.
Designed for deep reasoning, long context, and complex tool use.
Text Max: Frontier models for deep reasoning, long context, and complex workflows.
Highest minimum intelligence benchmark score across all Text instruments.
1M token context window and 128K token maximum output.
Accepts longer time to first token in exchange for frontier-class reasoning.
Designed for the most demanding workloads: extended analysis, multi-step synthesis, and complex tool use.
Which Instrument is right for you?
Choose Text Standard when:
You have tight latency budgets and need instant feeling UX.
You run many parallel calls where throughput and cost per token matter more than the last bit of reasoning quality.
Your prompts are short and you expect brief to moderate outputs.
You are doing routing, summarization, classification, or other transforms rather than complex planning.
Typical use cases:
Support chat assistants, help center bots, Slack and Discord helpers.
Email, meeting, and thread summarization, code diff explanations, document summaries.
RAG answers where context is already tight and responses should be concise.
Bulk text transforms such as classify, redact, extract fields, paraphrase, or translate short snippets.
Product surfaces that need instant feedback like autocomplete, inline suggestions, or hinting.
Choose Text Prime when:
You need higher reasoning quality for multi step analysis or synthesis across long context.
You expect very long outputs that must remain coherent.
You rely heavily on tool calling, retrieval, or other agent style workflows.
You are handling high stakes user facing answers where wrong but fast is not acceptable.
You prefer fewer escalations and are willing to pay more per Unit to get higher quality by default.
Typical use cases:
Deep code reasoning narratives like architecture reviews, migration plans, and extensive refactors with verification steps.
Executive assistants and research copilots that read large corpora and produce detailed briefs.
Strategic and planning agents for OKRs, roadmaps, project plans, and complex workflow orchestration.
Long form documents such as RFCs, memos, policy drafts, or multi chapter documentation.
High stakes customer facing answers where correctness is more important than raw speed or cost.
Choose Text Max when:
You need the absolute highest reasoning quality available on the market.
Your workload involves complex multi-step analysis, synthesis across very long context, or extended chain-of-thought reasoning.
You need frontier-class models for high-stakes decisions where the quality ceiling matters more than latency or cost.
You are building advanced agent or tool-use pipelines that require the most capable models to succeed.
Typical use cases:
Deep research and analysis: literature reviews, due diligence reports, multi-source synthesis across large document sets.
Complex agentic workflows that chain many reasoning and tool-use steps and cannot tolerate quality degradation.
Architecture-level code reasoning: full codebase analysis, cross-repo migration planning, and multi-file refactors with verification.
High-stakes professional outputs: legal analysis, financial modeling narratives, medical literature synthesis.
Any task where you would specifically choose the most capable model available (e.g., Claude Opus, GPT-5.4).
Last updated
Was this helpful?