Supported models
Choose the reasoning model that fits an agent's job, speed, and budget.
Hyperagent is model agnostic. You can choose models from multiple labs, change the model without rebuilding the agent, and give runtime helpers a different default from the parent.
Choose for the job
A model shapes reasoning quality, latency, and cost. Pick it after you understand the work:
How hard is the thinking?
Deep research, careful writing, and multi-step tool use benefit from a stronger model and often higher effort.
How often will it run?
Live Mode checks, high-volume replies, and simple classification reward speed and lower cost.
Who will act on the result?
External-facing work and consequential decisions deserve more quality budget than internal triage.
A long-lived agent should keep the model that matches its job. Do not let a one-off thread choice silently become the operating model for a recurring role.
Latest aliases
A Latest alias tracks the current stable model in a family. Use one when you want the agent to receive the family's normal upgrades without editing its configuration. Pin a numbered version when you need repeatability for an evaluation or controlled comparison.
The picker can show Latest aliases for Opus, Sonnet, Fable, Kimi, and GPT when that family is enabled for your workspace. It also offers numbered versions for users who want to stay put.
Anthropic models
| Model | Best fit | Context |
|---|---|---|
| Fable 5 | Highest-capability work when quality matters more than cost | 1M |
| Opus 5 | Complex agents, difficult reasoning, and careful final output | 1M |
| Opus 4.8 | Earlier flagship for complex work; supports Fast mode | 1M |
| Opus 4.7 | Earlier Opus for complex work | 1M |
| Opus 4.6 | Earlier Opus for complex work | 1M |
| Sonnet 5 | Balanced everyday agents at lower cost than Opus | 1M |
| Sonnet 4.6 | Earlier balanced Sonnet for everyday work | 1M |
| Haiku 4.5 | Quick, simple tasks and high-volume lightweight work | 200K |
OpenAI and Google models
| Model | Best fit | Context |
|---|---|---|
| GPT 5.6 Sol | OpenAI's most capable option for difficult agent work | 950K |
| GPT 5.6 Terra | Balanced OpenAI option for everyday agents | 950K |
| GPT 5.6 Luna | Fast, low-cost OpenAI option | 950K |
| GPT 5.5 | Earlier OpenAI flagship | 950K |
| Gemini 3.6 Flash | Fast structured execution, monitoring, research loops, and coding | 1M |
| Gemini 3.5 Flash | Earlier fast Google model for everyday work | 1M |
Open and specialist models
| Model | Best fit | Context |
|---|---|---|
| Kimi K3 | Moonshot's highest-capability open model | 900K |
| Kimi K3 Fast | Faster Kimi K3 output when latency matters | 900K |
| Kimi K2.6 | Low-cost everyday work on an open model | 230K |
| GLM 5.2 | Low-cost everyday agent work | 1M |
| GLM 5.2 Fast | Faster GLM output for high-volume work | 1M |
| Qwen 3.7 Plus | Cost-aware general work and coding | 230K |
| DeepSeek V4 Pro | Deep reasoning on an open model | 1M |
| Fugu Ultra | Coordinated expert-agent work on complex tasks | 1M |
| Grok 4.5 | Conversational agent work | 500K |
| Muse Spark 1.1 | Creative agent work | 1M |
| Inkling | Compact, low-cost open-model work | 230K |
The catalog changes as Hyperagent validates additional models. Trust the picker for the list available to you. If a model you need is missing, share the use case with the Hyperagent team so the request includes the job, not only the model name.
Thinking, speed, and budget
The model is only one part of the cost and quality decision. Agent settings can also expose effort, a per-run budget limit, and a default model for runtime helpers.
Effort
Many models support an effort control. Higher effort gives the model more room for difficult reasoning and usually increases latency and cost.
| Job | Starting point |
|---|---|
| Hard analysis, debugging, or strategy | Stronger model, medium or high effort |
| Routine classification or simple rewrites | Faster model, lower effort |
| Live Mode heartbeat | Capable low-cost model, lower effort, precise checklist |
Fast mode
Fast mode runs the same eligible model at higher throughput. Use it for interactive replies and tight loops where latency matters. Standard mode is usually the better fit for deep analysis and final writing.
If you do not see a toggle for fast mode on the model you're using, the model does not expose Fast mode in Hyperagent.
Budget limit per run
A per-run budget limits how much one agent query may spend. Use it for schedules, Live Mode, and agents that browse or loop through large workloads. A cap is a guardrail; it does not replace picking a sensible model and effort level.
Recipes
| Agent job | Model direction | Effort | Cost posture |
|---|---|---|---|
| Primary workhorse | Latest Opus or Sonnet | Medium to high | Spend on judgment |
| Live Mode watch | Sonnet, Haiku, or another fast tier | Low | Optimize recurring checks |
| High-volume simple work | Haiku or a low-cost alternative | Low | Keep each run lean |
| Deep research specialist | Flagship or deep-reasoning model | High | Tight tools and clear stop conditions |
| Early experiment | Faster tier first | Low to medium | Promote only where quality fails |