Skip to content

AI models

TeXRA connects directly to frontier reasoning models from leading providers—including Anthropic, OpenAI, Google, DeepSeek, Moonshot, and DashScope. You can assign flagship reasoning models to demanding mathematical proofs and autonomous loops, while reserving faster, cost-efficient variants for routine document polish and formatting. Select a model from the dropdown menu in the TeXRA UI or pass --model in the CLI. Hover over any option to inspect its context window and pricing.

Model
sonnet5T
opus5TT
opus5
sonnet5TT
sonnet5
haiku45TT
haiku45

The model picker: monospace model ids with a T badge on thinking variants, and a hover popover showing context window and per-1M token pricing.

Model ID suffixes:

  • T = thinking/reasoning mode enabled (shows chain-of-thought)
  • - = lighter, faster variant
  • Numbers indicate the version (for example, 45 = 4.5, 25 = 2.5)

Anthropic models

Model IDUse CaseCostSpeed
fable51Most capable, always-on adaptive thinking$$$$Slow
opus5TTop-tier reasoning, long-horizon work$$$$Slow
opus5Most capable for agentic coding$$$$Slow
sonnet5TAll-rounder with reasoning$$$Medium
sonnet5Strong all-rounder$$$Medium
haiku45TFast with reasoning$$Fast
haiku45Fast responses$$Fast

Fable 5.1, Opus 5, and Sonnet 5 (and the older Opus 4.6 to 4.8 and Sonnet 4.6) include the full 1M context window at standard pricing, with no opt-in or beta header required. Haiku 4.5, Opus 4.5, and Sonnet 4.5 use a 200K context window.

Claude Fable 5.1 (fable51) is Anthropic's most capable model. Thinking is always on (adaptive, with summarized reasoning), so there is no separate T variant. It supports the full reasoning-effort range up to Extra High and the top Max tier, and is eligible for context compaction in tool-use mode.

Claude Opus 5 uses adaptive thinking only (extended thinking with a manual budget_tokens is no longer accepted). TeXRA's reasoning-effort selector maps to Anthropic's effort levels automatically: pick opus5T with Extra High (or the top Max tier) effort for the strongest agentic coding and long-horizon tasks. Opus 5 also supports high-resolution images for better figure, chart, and screenshot understanding. TeXRA downscales images above texra.maxImageDimension (default 2000px) before sending, so raise that setting to send higher-resolution figures.

OpenAI models

Model IDUse CaseCostSpeed
gpt6Most capable, 1M context$$$$Fast
gpt56proPro reasoning mode, 1M context$$$$Slow
gpt56Flagship reasoning, 1M context$$$$Medium
gpt56fastFlagship, fast variant$$$$Fast
gpt56-Lower-cost reasoning$$$Fast
gpt56--Budget reasoning$Fast

GPT-6 Astra (gpt6) is OpenAI's most capable model for the hardest end-to-end work; it is available in the API and in Codex for Pro, Enterprise, and Business Premium subscribers, and supports reasoning effort up to Max. Note its long-context pricing: prompts above 272K input tokens bill at 2x input/cache and 1.5x output for the full request.

GPT-5.6 Sol (gpt56) is OpenAI's current flagship reasoning model; TeXRA pins the Codex integration to gpt-5.5. GPT-5.6 Pro (gpt56pro) runs the same model in the Responses API's pro reasoning mode, billed at standard token rates rather than a premium tier, for the hardest planning and long-horizon tasks. It is hidden by default; enable it from Settings → Providers & Models when you need it. For one-off hard questions you can also enable the inquiry tool and paste the answer from your own ChatGPT subscription instead of running a full agent turn against the API. gpt56- (Terra) and gpt56-- (Luna) are the lower-cost options for most workloads. Read the OpenAI API reference for full capabilities.

GPT-5 reasoning summaries require account verification. Enable them with texra.model.gpt5ReasoningSummary.

Google models

Model IDUse CaseCostSpeed
gemini31pPro with reasoning, 1M context$$$Medium
gemini38fFlash model with 1M context$$Fast

DeepSeek models

Model IDUse CaseCostSpeed
deepseek41V4.1 Flash chat mode$Fast
deepseek41TV4.1 Flash with reasoning$Medium
deepseekV4 Flash chat mode$Fast
deepseekTV4 Flash with reasoning$Medium
deepseekvisionV4 Flash Vision (Exp), image input$Fast
deepseekproV4 Pro chat mode$Medium
deepseekproTV4 Pro with reasoning$Medium

Moonshot Kimi models

Model IDUse CaseCostSpeed
kimi3K3 flagship, 1M context$$$Medium

DashScope Qwen models

Model IDUse CaseCostSpeed
qwenplusHybrid thinking, 1M context$$Medium
qwenturboFast with optional thinking$Fast

MiniMax models

Model IDUse CaseCostSpeed
minimaxM3Flagship with interleaved thinking, 1M context$Medium

MiniMax uses interleaved thinking (chain-of-thought woven into responses). API keys are region-specific: international keys (api.minimax.io) and China keys (api.minimaxi.com) are not interchangeable. Expand the MiniMax row in Providers & Models → API configuration and toggle MiniMax China region (GLM, Kimi/Moonshot, and Qwen have matching toggles; GLM's is on by default).

GLM (Zhipu AI / Z.AI) models

Model IDUse CaseCostSpeed
glm53Flagship, 1M context, reasoning-effort tiers$$Medium
glm53flashFlagship fast variant, 1M context$Fast
glm52Previous flagship (deprecated)$$Medium
glm5turboFast inference, agent-optimized$$$Medium
glm5vturboMultimodal vision model$$Medium

GLM models support thinking mode (reasoning is shown inline). The API uses a non-standard base path (/api/paas/v4), which TeXRA handles automatically.

  • International (Z.AI): Get your API key at z.ai; endpoint: api.z.ai
  • China (BigModel): Get your API key at open.bigmodel.cn; endpoint: open.bigmodel.cn (default)
  • Coding Plan: GLM offers monthly subscription plans as an alternative to pay-as-you-go, with access to all GLM models. Coding Plan uses a separate endpoint (/api/coding/paas/v4). Turn on the Coding Plan toggle in the Providers & Models tab. Subscribe to the GLM Coding Plan.

Meta (Muse Spark) models

Model IDUse CaseCostSpeed
musespark13Flagship reasoning + vision, 1M context$$Medium
musespark11Reasoning + vision + PDF, 1M context$$Medium

Muse Spark always reasons (effort is adjustable but cannot be disabled). TeXRA uses the Meta Model API's Responses surface, which carries reasoning across turns and supports tool calling. The API is in public preview for US-based developers.

  • Get your API key at dev.meta.ai (Model API dashboard → API keys tab)

Grok / xAI models

Model IDUse CaseCostSpeed
grok45Reasoning + vision$$$Medium

Choosing a model

Simple tasksFast, cheap models
gpt56--deepseekhaiku45
Complex tasksPowerful flagship models
fable51opus5gpt56gemini31p
Code-heavy / LaTeX editingStrong editing models
opus5Tsonnet5Tgpt56
Reasoning-heavyThinking models
fable51opus5Tsonnet5TdeepseekTkimi3
Large documentsHigh-context models
gemini31pfable51sonnet5opus5

Pick a model by intent: each use case maps to a short list of recommended model ids.

Setting API keys

Subscription-backed models in VS Code

The VS Code extension can also use compatible models from a GitHub Copilot subscription. Open Settings → Subscriptions → Copilot in VS Code, then select Grant access. VS Code shows its own consent prompt; TeXRA never asks for or stores a Copilot API key.

Copilot models appear only in the VS Code extension because the official Language Model API is an editor capability. They do not appear in the CLI or desktop model lists. If Copilot quota is exhausted, the retry panel can start a new run through the corresponding provider model once a usable provider API key is available.

Using your own provider API key? TeXRA stores keys in VS Code's secret storage; they are never written to settings files.

  1. Open the Settings Dashboard: Select the gear icon at the top of the TeXRA panel, or run TeXRA: Show Settings Dashboard from the Command Palette.
  2. Go to the Providers & Models tab: The API configuration table lists every provider with its current key status (Set, Env, or Not set).
  3. Set the key: Find your provider's row and select the Set API key button, then paste your key. If you don't have a key yet, select the Get button to open the provider's API key page.

The Status column shows Set once the key is stored. To replace a key, set it again; to remove one, select the trash icon. Repeat for each provider you plan to use.

Dashboard — texra-paper

API Configuration

Chat subscriptions (ChatGPT Plus, Claude Pro, etc.) do not include API access — you need a key from the provider's developer platform.

ProviderStatusActions
Anthropic
Key set
OpenAI
Key set
Google
Env
xAI
Not set
DeepSeek
Not set

The Providers & Models tab's API configuration table: each provider shows its key status and Set / Get / Remove actions.

Per-provider settings

Expand a provider's row (select the chevron) to point requests at a custom endpoint, for providers that support it.

You can also place a .env file in your workspace with variables such as OPENAI_API_KEY. TeXRA loads it automatically, so you don't need to enter keys each time.

Already paying for a ChatGPT or Grok subscription? Sign in and skip the API key for those models. Kimi Code and the GLM Coding Plan also run on a subscription you already pay for, authenticated with a plan-specific key instead of a full provider key. Read Quick start → Add a key or connect a subscription.

Customizing the model list

Choose which models appear in the extension picker from the Dashboard → Providers & Models tab: toggle them on or off per provider, no JSON required (the choice is saved in the extension).

In the CLI TUI, run /model after a chat starts to see the models your current credentials can run. Mid-session switching is limited to models that share the active model's provider family; other entries are shown disabled with a reason, and switching waits until the current response finishes. To change family, start a new chat with --model. Before you send the first message, /agent chains straight into that same model picker, so choosing a root agent and its model stays one step.

For headless CLI runs, list what is available with texra models list (or texra models show <id> for details), then pick a default for your project by setting the model key in .texra/config.json, or override per run with --model <id>:

texra models
$texra models list
fable51Claude Fable 5.1api key set
opus5TOpus 5 (Thinking)api key set
sonnet5TSonnet 5 (Thinking)api key set
deepseekTDeepSeek V4 Flash (Thinking)api key set
$texra models show fable51
id: fable51
label: Claude Fable 5.1
provider: anthropic
status: api key set

The id column is exactly what --model takes: the same short ids used in the tables above. --all includes models your current credentials can't run, with the reason.

Using OpenRouter

To access additional models or alternative pricing:

  1. Get an OpenRouter API key
  2. Add it with the TeXRA: Set API Key command
  3. In the Dashboard → Providers & Models tab → API configuration, expand the OpenRouter row and turn on Use OpenRouter for all models

Expanding any provider's row in API Configuration reveals its key field plus the per-provider toggles described here:

OpenAI
Key set
sk-••••••••••••••••••••••••
Route this provider's calls through OpenRouter

Expand a provider's API configuration row to reveal its masked key field (and Custom endpoint where supported); the OpenRouter row adds Use OpenRouter for all models.

Next steps