Core References

AI Model Reference Guide

Compare current AI model families, understand reasoning and sampling controls, and choose models for research reasoning, writing, and analysis tasks

Reference page for: AI model selection and configuration. Other pages link here for current model-family characteristics and provider-specific settings.

Overview: Current, but Not Permanent

Warning

AI model names, aliases, pricing, rate limits, and control parameters change quickly. This page was spot-checked against provider documentation on June 23, 2026. Before building a workflow around a specific model ID, check the provider's current model list.

The practical question is not "which model is best?" It is which model, access path, and control surface fit the task:

Reasoning depth: Can it handle theoretical tension and multi-step analysis?
Context capacity: Can it read the relevant corpus without destructive chunking?
Writing behavior: Does it preserve academic nuance and voice?
Cost and latency: Can you afford to iterate?
Control surface: Does it want reasoning effort, thinking level, prompt constraints, or sampling parameters?

Which AI Model Should I Use?

Best starting points:

Claude Opus 4.8 or Claude Sonnet 4.6 for theory, critique, and nuanced writing.
GPT-5.5 for reliable complex reasoning, coding-adjacent work, and structured fixes.
Gemini 3.5 Flash or Gemini 3.1 Pro preview for large-context synthesis.
GLM-5.2 for long-context coding-agent reasoning and Chinese-English synthesis.

Use cases: theory building, critical analysis, synthesis, methodology design.

Pro tip: Start with cheaper or free access for exploration, then move the strongest prompt and evidence set to the model you trust for final reasoning.

Full comparison: AI Model Reference Guide

Current Model Families

OpenAI GPT Family

Current anchor: GPT-5.5 for complex reasoning and coding; GPT-5.4, GPT-5.4 mini, and GPT-5.4 nano for lower cost/latency.
Context: up to 1M on GPT-5.5 / GPT-5.4; smaller variants vary.
Control surface: reasoning.effort where supported; sampling defaults are usually enough unless a specific API task needs tuning.
Best for: reliable editing, coding-adjacent research infrastructure, broad general reasoning.
Source: OpenAI model docs

Anthropic Claude Family

Current anchor: Claude Opus 4.8, Claude Sonnet 4.6, Claude Haiku 4.5.
Special status: Claude Mythos Preview is invitation-only and focused on defensive cybersecurity workflows. Claude Fable 5 launched June 9, 2026 as Anthropic's first public Mythos-class model, then was suspended globally on June 12, 2026 following a US export-control directive; do not rely on it until access is restored.
Control surface: effort for Opus 4.8; do not set non-default temperature, top_p, or top_k on Opus 4.8 Messages API requests.
Best for: theoretical writing, qualitative interpretation, long-form critique, careful synthesis.
Source: Claude model docs

Google Gemini Family

Current anchor: Gemini 3.5 Flash stable; Gemini 3.1 Pro preview; Gemini 3 Flash preview; Gemini 3.1 Flash-Lite stable.
Context: Gemini remains the large-context workhorse; exact limits and rate limits vary by tier and model.
Control surface: for Gemini 3.x, remove temperature, top_p, and top_k; use thinking_level where relevant.
Best for: large-corpus synthesis, volume processing, multimodal tasks, Google ecosystem workflows.
Source: Gemini API model docs

DeepSeek Family

Current anchor: DeepSeek V4-era models, including deepseek-v4-flash.
Alias caveat: deepseek-chat and deepseek-reasoner are compatibility aliases scheduled for deprecation on July 24, 2026 at 15:59 UTC.
Control surface: temperature is still documented by use case; default is 1.0, with coding/math at 0.0, analysis at 1.0, and creative work higher.
Best for: high-volume exploration, budget-sensitive screening, reasoning checks where cost matters.
Source: DeepSeek pricing/model docs

Kimi Family (Moonshot AI)

Current anchor: kimi-k2.7-code, kimi-k2.7-code-highspeed, kimi-k2.6, and kimi-k2.5.
Context: 256K for current K2 models.
Control surface: do not set temperature for K2.7 Code or K2.6; use the default. K2.7 Code keeps thinking on, while K2.6 supports thinking/non-thinking modes.
Best for: code review, adversarial challenge, long-document critique, agentic workflows.
Source: Kimi model docs

GLM Family (Z.ai / Zhipu AI)

Current anchor: GLM-5.2.
Context/output: 1M context and up to 128K output.
Control surface: reasoning_effort is supported by GLM-5.2; default temperature for GLM-5.2 is 1.0 if you set sampling at all.
Best for: long-context coding-adjacent reasoning, multilingual Chinese-English synthesis, frontier-adjacent open-weight experiments.
Source: GLM-5.2 docs

Qwen Family (Alibaba)

Current anchor: Qwen3.5-Plus and Qwen3.5-Flash in global/international modes, with Qwen3-Max as the complex-task flagship where available.
Context: Plus/Flash entries list 1M context in supported deployment modes; Max is smaller but more capable for difficult tasks.
Lineup note: Qwen3.5-Plus is not a successor to Qwen3-Max. Plus is the large-context, cost-efficient workhorse; Max is the smaller-context flagship for harder reasoning. Pick by task, not by version number.
Caveat: model names and availability differ by international, global, US, EU, Hong Kong, and Chinese mainland deployment modes.
Best for: multilingual work, open-source comparisons, region-specific provider setups.
Source: Alibaba Model Studio model list

xAI Grok Family

Current anchor: Grok 4.3 for general text work; Grok Build 0.1 for coding-focused work.
Context: Grok 4.3 lists a 1M-token context window.
Control surface: reasoning_effort on Grok 4.3 defaults to low; some penalty/stop parameters cannot be used with reasoning models.
Best for: fast agentic tool calling, broad general tasks, xAI ecosystem experiments.
Source: xAI model docs

Local / Open-Weight Models

Current anchor: depends on your runtime, hardware, and privacy constraints.
Best for: sensitive data, reproducibility, offline work, customization.
Caveat: local quality depends more on deployment details than provider marketing pages.

Access Paths

Access path	Best for	Notes
Provider web app	Fast exploration	Good for trying a model before wiring API access.
Google AI Studio	Gemini API keys and free-tier experiments	Use for direct Gemini model access, not Antigravity CLI auth.
Cherry Studio	GUI model comparison and knowledge bases	Best first interface for many researchers.
OpenCode	CLI-side provider comparison	Useful when model choice is part of the method.
Vox MCP	Multi-model access inside an MCP client	Pure passthrough; good for challenge and triangulation.
Claude Code / Antigravity CLI	Agentic project work	Product sign-in flows, not generic provider-key clients.

Configuration: Defaults First

Sampling is no longer the default lever

The old advice was to tune temperature for every model. That is now misleading. Many current reasoning models either reject sampling changes, ignore them, or explicitly recommend defaults.

Default rule: omit temperature, top_p, and top_k unless the provider currently documents that the model supports and benefits from changing them.

Provider-specific notes:

Provider/model	Recommendation
Claude Opus 4.8	Omit non-default `temperature`, `top_p`, `top_k`; non-default values can error.
Gemini 3.x	Remove `temperature`, `top_p`, `top_k`; use `thinking_level` and prompt constraints.
Kimi K2.7 / K2.6	Do not set `temperature`; it is not modifiable on these models.
GLM-5.2	Default temperature is `1.0`; prefer `reasoning_effort` for thinking control.
DeepSeek	Temperature remains documented by use case; default is `1.0`.
xAI Grok 4.3	Prefer `reasoning_effort`; watch incompatible parameters on reasoning models.

Tip

If a compatible model forces you to choose a temperature and the provider does not give task-specific guidance, start at 1.0. Drop lower only for mechanical extraction or classification where repeatability matters.

Prefer reasoning controls when available

Control	Providers/examples	Use for
`reasoning.effort` / `effort`	OpenAI, Claude, xAI, GLM variants	More or less deliberate reasoning.
`thinking_level`	Gemini 3.x	Gemini reasoning depth without raw sampling changes.
thinking mode	Kimi K2.6, DeepSeek V4-era modes, GLM	Choosing visible/hidden thinking behavior where supported.
Sequential Thinking MCP	Any MCP-capable client	Tool-visible, revisable reasoning steps.

Use high reasoning effort for theoretical integration, methods design, and hard debugging. Use low/default effort for routine extraction, tagging, and quick checks.

Task type	Preferred control	Suggested starting point
Theory synthesis	Reasoning effort / thinking level	Medium or high; keep sampling defaults.
Framework building	Reasoning effort plus explicit criteria	High; ask for alternatives and tensions.
Literature synthesis	Context capacity plus structured prompt	Medium/high thinking; defaults for sampling.
Draft writing	Prompt constraints and examples	Defaults first; ask for voice and structure directly.
Citation extraction	Deterministic prompt rules	Low reasoning or provider-specific low temp if supported.
Systematic coding	Rubric and validation pass	Low/medium reasoning; prioritize consistency checks.

Strategic Model Usage

Use cheap or free access paths to learn model behavior against your own corpus.

Start with Google AI Studio, Cherry Studio, OpenCode, or OpenRouter.
Run the same prompt across at least three model families.
Record failure modes, not just best answers.
Promote a model only after it handles your real material.

Analysis type	Good starting models	Why
Large literature sets	Gemini 3.5 Flash, Gemini 3.1 Pro preview, GLM-5.2	Large context and long-horizon reasoning.
Theoretical depth	Claude Opus 4.8, Claude Sonnet 4.6	Strong interpretive and argumentative behavior.
Critical challenge	Kimi K2.7 / K2.6, Claude Opus 4.8	Useful for adversarial pressure and blind-spot detection.
Consistent coding/repo work	GPT-5.5, Claude Sonnet 4.6, GLM-5.2	Strong project reasoning and implementation discipline.
Budget-conscious screening	DeepSeek V4-era models, Gemini Flash/Lite, Qwen3.5-Flash	Cheap enough to iterate.

Free and Low-Cost Options

Google AI Studio remains the easiest free/low-cost backup for Gemini API work, but limits change often. Check Gemini rate limits before relying on a quota for teaching or batch work.

Other cost-management paths:

Use OpenCode or Cherry Studio to compare providers before committing to a long run.
Use DeepSeek or Qwen for high-volume screening when quality is sufficient.
Use local models when privacy matters more than frontier quality.
Reserve expensive models for the final synthesis, argument, or validation pass.

Next Step

The reference above tells you what is available. The companion page tells you how to test those choices against your own materials: AI Model Discovery Protocol.