Technical referenceOfficial documentation5 min readreference-data

FastGPT LLM Capability Tiers Reference

The 268 chat models bundled with fastgpt-plugin v1.1.3, grouped by context length and capability fields

This page groups the 268 chat models bundled with fastgpt-plugin v1.1.3 by context length, maximum single response, quote limit, image input and tool calling. Models in the same tier have identical values across these fields and can be configured with the same settings.

Overview

ItemCount
Bundled chat models268
Providers21
Distinct capability tiers (provider + fields)124
Marked as supporting tool calling210
Marked as supporting image input135

Distribution by context length

Context lengthModels
1M and above69
200K–1M66
100K–200K71
32K–100K38
Below 32K24

Distribution by provider

ProviderModelsCapability tiers
Qwen3813
OpenAI328
ChatGLM2710
StepFun1613
Claude153
AntLing148
Ernie1410
Gemini143
Hunyuan1411
MiniMax116
Moonshot118
MistralAI104
Doubao93
Groq94
Baichuan83
Grok73
SparkDesk74
DeepSeek54
Siliconflow33
InternLM21
Yi22

Three things to keep in mind

  1. Capability fields are the values registered in the bundled catalogue. Actual availability also depends on whether the connected channel exposes that capability.
  2. Context length (maxContext) caps how much retrieved content fits in one request. The quote limit (quoteMaxToken) is a token budget for cited content — it caps the total tokens quotes may occupy, not the number of segments; segment count is set on the retrieval side. Read them together; looking at only one leads to a wrong recall setting.
  3. Models in the same tier share these field values, which does not mean they produce answers of the same quality.

Scope of this reference

Content is taken from the bundled model catalogue in fastgpt-plugin v1.1.3. The following are out of scope:

  • Custom models registered by the operator
  • Quota, region and version differences on the provider side
  • Capability fields not present in the bundled catalogue

References