| title | Models | ||
|---|---|---|---|
| description | Every registered model id, its context window, its reasoning variants and its list-price multiplier. | ||
| sidebar |
|
The plugin auto-registers the following, and they appear in the model picker with no extra config: Haiku 4.5/5.5, Sonnet 4.6/5/5.5, Opus 4.5/4.6/4.7/4.8/5/5.5 (plus fast-mode entries for Opus 5 and 5.5), Fable 5/5.1/5.2 and Mythos 5/5.1/5.2 (plus a fast-mode entry for each 5.2). Each carries the reasoning variants Claude Code accepts for it: none on Haiku 4.5, fewer on Opus 4.5 and the 4.6 models, and all five (low / medium / high / xhigh / max) on everything newer. By default the list, the windows and the variants follow what your installed claude reports; see Learned from the Claude CLI.
| ID | Display name | Context | Output | Reasoning variants | Price × |
|---|---|---|---|---|---|
claude-haiku-4-5 |
Claude Haiku 4.5 | 200k | 64,000 | – | 1× |
claude-haiku-5-5 |
Claude Haiku 5.5 | 1M | 128,000 | low/medium/high/xhigh/max | 0.1× |
claude-sonnet-4-6 |
Claude Sonnet 4.6 | 200k | 128,000 | low/medium/high/max | 3× |
claude-sonnet-5 |
Claude Sonnet 5 | 1M | 128,000 | low/medium/high/xhigh/max | 2× |
claude-sonnet-5-5 |
Claude Sonnet 5.5 | 1M | 128,000 | low/medium/high/xhigh/max | 2× |
claude-opus-4-5 |
Claude Opus 4.5 | 200k | 64,000 | low/medium/high | 5× |
claude-opus-4-6 |
Claude Opus 4.6 | 200k | 128,000 | low/medium/high/max | 5× |
claude-opus-4-7 |
Claude Opus 4.7 | 1M | 128,000 | low/medium/high/xhigh/max | 5× |
claude-opus-4-8 |
Claude Opus 4.8 | 1M | 128,000 | low/medium/high/xhigh/max | 5× |
claude-opus-5 |
Claude Opus 5 | 1M | 128,000 | low/medium/high/xhigh/max | 5× |
claude-opus-5-fast |
Claude Opus 5 Fast | 1M | 128,000 | low/medium/high/xhigh/max | 10× |
claude-opus-5-5 |
Claude Opus 5.5 | 1M | 128,000 | low/medium/high/xhigh/max | 4× |
claude-opus-5-5-fast |
Claude Opus 5.5 Fast | 1M | 128,000 | low/medium/high/xhigh/max | 8× |
claude-fable-5 |
Claude Fable 5 | 1M | 128,000 | low/medium/high/xhigh/max | 10× |
claude-fable-5-1 |
Claude Fable 5.1 | 1M | 128,000 | low/medium/high/xhigh/max | 10× |
claude-fable-5-2 |
Claude Fable 5.2 | 1M | 128,000 | low/medium/high/xhigh/max | 8× |
claude-fable-5-2-fast |
Claude Fable 5.2 Fast | 1M | 128,000 | low/medium/high/xhigh/max | 16× |
claude-mythos-5 |
Claude Mythos 5 | 1M | 128,000 | low/medium/high/xhigh/max | 10× |
claude-mythos-5-1 |
Claude Mythos 5.1 | 1M | 128,000 | low/medium/high/xhigh/max | 10× |
claude-mythos-5-2 |
Claude Mythos 5.2 | 1M | 128,000 | low/medium/high/xhigh/max | 8× |
claude-mythos-5-2-fast |
Claude Mythos 5.2 Fast | 1M | 128,000 | low/medium/high/xhigh/max | 16× |
claude-mythos-5 and claude-mythos-5-1 are Mythos-class counterparts to the corresponding Fable models, but without safety classifiers, and are limited availability via Project Glasswing. They're registered unconditionally; if your Claude account lacks access, claude --model just errors. Use the corresponding generally available claude-fable-5 or claude-fable-5-1 otherwise. claude-mythos-5-2 is the same model as claude-fable-5-2, available only to organizations verified through Anthropic's verification programs, such as the Cyber Verification Program.
Capabilities for every model: text, image and PDF input, text output, tool use, attachments. No temperature control, no audio/video, no interleaved streaming.
Price × is each model's per-token list price relative to Haiku 4.5, which anchors the scale at 1×. It's derived exactly from Anthropic's published pricing (input and output ratios both come out the same: Haiku 4.5 $1/$5 = 1×, Haiku 5.5 $0.10/$0.50 = 0.1×, Sonnet 5 and 5.5 $2/$10 = 2×, Sonnet 4.6 $3/$15 = 3×, Opus 5.5 $4/$20 = 4×, Opus $5/$25 = 5×, Opus 5.5 fast mode and Fable/Mythos 5.2 $8/$40 = 8×, Fable/Mythos 5 and 5.1 / Opus 5 fast mode $10/$50 = 10×, Fable/Mythos 5.2 fast mode $16/$80 = 16×). So Fable/Mythos 5 and 5.1, and fast-mode Opus 5, all cost 2× standard Opus 5, and fast mode is 2× the standard price on every model that offers it. The same multiplier is shown as a (N×) suffix on the display name in opencode's model picker, since opencode has no dedicated multiplier field. On a flat Max/Pro subscription it doubles as a rough guide to how fast each model drains your usage limit.
Fable 5.1 and Mythos 5.1 keep the same $10/M input and $50/M output rates as 5.0, but cache reads cost $0.25/M instead of $1/M. Their cache-write rate remains $12.50/M.
Fable 5.2 and Mythos 5.2 (released 2026-11-09) are the first Fable/Mythos models below the $10/$50 line: $8/M input and $40/M output, with cache reads at $0.20/M (the same 0.025x input ratio as 5.1) and cache writes at $10/M (1-hour writes $16/M). They need Claude Code 2.1.321 or newer, the release that added them and made Fable 5.2 the default Fable; an older CLI still serves them, on fallback limits (a 200k window), and /claude-code-doctor lists them as models it does not know. Both have a fast mode at $16/M input and $80/M output, registered as the -fast entries below.
Haiku 5.5 (released 2026-10-07) is a tenth of Haiku 4.5 on every axis: $0.10/M input, $0.50/M output, cache writes at $0.125/M and cache reads at $0.01/M, with a 1M context window, 128,000 output tokens and adaptive thinking, so unlike Haiku 4.5 it carries the reasoning variants. It needs Claude Code 2.1.293 or newer, the release that added it and made it the default Haiku; an older CLI still serves it, on fallback limits, and the plugin warns once.
Haiku 5.5 is also the one model Anthropic prices by prompt length: a request whose prompt (input plus cached read plus cache write) is over 100,000 tokens bills every one of its tokens at five times the rate, $0.50/M input and $2.50/M output. The plugin applies that to the per-turn cost it reports on the interactive transport, where it rebuilds the figure itself. The Price × column and the cost opencode computes from the model catalog are the base (up to 100,000 tokens) rate: opencode's config schema has only a fixed 200,000-token tier field, which cannot express a 100,000-token threshold, so filling it in would be wrong between 100k and 200k while implying the tier was modelled. Budget for up to 5× the quoted price on long-prompt Haiku 5.5 turns.
Sonnet 5 and Sonnet 5.5 are $2/M input and $10/M output, with cache writes at $2.50/M (1-hour writes $4/M). Cache reads are $0.20/M on Sonnet 5 and $0.10/M on Sonnet 5.5 (0.05× input, like Opus 5.5); earlier releases of the plugin registered Sonnet 5.5 at $0.20/M, so the cost opencode showed for its cache reads was twice the bill. Sonnet 5's price was announced as introductory until 2026-08-31, but Anthropic cancelled the increase to $3/$15, so $2/$10 is its standard price. Sonnet 5.5 runs on any recent Claude Code, but 2.1.284 is the first release that knows it. An older CLI still serves it, on fallback limits (a 200k context window instead of 1M, and an estimated cost). The plugin warns once when the CLI reports that, and claude update fixes it.
claude-sonnet-4-5 is retired. Anthropic retires Sonnet 4.5 on the Claude API on 2026-11-30, and after that a request to it fails (the model deprecations page names claude-sonnet-5-5 as the replacement). Claude Code does not remap it: its own remap of retired models covers Opus 4 and 4.1 only, and any other retired id is a "model not found" error. So the plugin no longer offers it in the picker, and an id still named where the plugin resolves models itself (forceModel, defaultSubagentModel, fallbackModels, compactionModel, titleModel, or a model you registered under provider.claude-code.models, the dated claude-sonnet-4-5-20250929 included) runs claude-sonnet-5-5 instead, from the day you upgrade, with one warning per opencode process naming the replacement and both prices: $2/$10 per million input/output tokens, against Sonnet 4.5's $3/$15. Sonnet 5.5 takes reasoning variants and has a 1M window; a model you registered yourself keeps the window you gave it. A model opencode resolves itself (an agent's model:, the top-level model) is refused by opencode as unknown, so change those to claude-sonnet-5-5.
Opus 5.5 is priced below the Opus line at $4/M input and $20/M output, with cache writes at $5/M and cache reads at $0.20/M (0.05× input rather than the usual 0.1×). It needs Claude Code 2.1.280 or newer: the API rejects it from an older CLI with a 400 naming that floor, which shows up as a failed turn.
The model ID is passed straight through to claude --model, so anything Claude Code accepts works. The -fast IDs are the one exception, described below.
claude-opus-5-5-fast, claude-opus-5-fast, claude-fable-5-2-fast and claude-mythos-5-2-fast run the same models at up to 2.5× the output tokens per second, at 2× the price ($8/M input, $40/M output for Opus 5.5, the 8× column; $10/M input, $50/M output for Opus 5, the 10× column; $16/M input, $80/M output for Fable and Mythos 5.2, the 16× column). Pick them in the model selector like any other model. The two 5.2 entries need Claude Code 2.1.321 or newer: an older CLI does not know those models, refuses them fast mode, and the turn runs at standard speed with the warning below.
claude-opus-4-8-fast is retired. Anthropic deprecated Opus 4.8's fast mode on 2026-10-30 and removes it on 2026-11-30; after that a fast request on Opus 4.8 is an error, with no fallback to standard speed (Claude Code 2.1.316's changelog). Claude Code itself does nothing about it, so the plugin no longer offers the id in the picker, and an id still named where the plugin resolves models itself (forceModel, defaultSubagentModel, fallbackModels, or a model you registered under provider.claude-code.models) runs plain claude-opus-4-8 at standard speed and the standard price, with one warning per opencode process naming claude-opus-5-5-fast as the replacement. A model opencode resolves itself (an agent's model:, the top-level model) is refused by opencode as unknown, so change those to claude-opus-5-5-fast (or claude-opus-4-8). Opus 5's fast mode, a research preview, is unaffected.
The -fast suffix is this plugin's own marker, not a model name Anthropic serves. The plugin strips it and spawns claude --model claude-opus-5 --settings '{"fastMode":true}', because that settings layer is the only way to opt a headless (--print) session into fast mode: there is no --fast flag, and the old claude-opus-4-6-fast style model names are retired. Requires Claude Code 2.1.220+; below that the plugin skips the opt-in and you get standard speed.
Fast mode is not available everywhere, and it fails soft: an ineligible account drops back to standard speed with no error. Known blockers:
- Usage credits are off. The most common one. Run
/usage-creditsin an interactiveclaudesession to enable them. - Not first-party. Fast mode is Anthropic-API-only; Bedrock, Vertex, and Foundry are excluded.
- Free tier, or an organization that has turned fast mode off.
- Cooldown. Fast mode has its own rate limit; after a hit, Claude Code falls back to standard until it clears.
CLAUDE_CODE_DISABLE_FAST_MODE=1in the environment turns it off outright.
Because a downgrade is otherwise invisible, and because the picker shows these IDs at their fast price regardless, the plugin logs a warning (once per reason) when a fast turn actually ran at standard speed, naming the reason. If you see it, switch to the non-fast ID so the picker's price matches your bill.
The warning is read off the end of the turn, not the start. Claude Code answers the question twice and the two can disagree: measured on 2.1.288, an account with usage credits off is told fast_mode_state: "on" when the session opens and "off" with extra_usage_disabled on the terminal result, after the turn has already run and billed at standard Opus rates. Until v0.48 the plugin read only the first one, so for that account it logged "fast mode active" and never warned at all.
The table above is the plugin's own list. With cliModelCatalog on (the default), what opencode registers follows your installed Claude Code instead, wherever the two disagree:
- Context window: the window the CLI actually runs the model in. Sonnet 4.6 and Opus 4.6 are the measured case: Anthropic's API offers them a 1M window, but
claude --model claude-opus-4-6runs in 200k (Claude Code 2.1.288 to 2.1.314), and the 1M window is the separateclaude-opus-4-6[1m]id. Telling opencode 1M meant its context gauge and its own compaction never saw the limit the CLI was working to. - Reasoning variants: only the effort levels the CLI accepts for that model. Claude Code quietly runs a level a model lacks as
high, so a variant it does not accept is a choice in the picker that changes nothing. - New models: a model the CLI's own picker offers that this plugin does not list yet is added, named the way the CLI names it. Its price is unknown, so it is registered at zero, like any model you add yourself;
forceModelanddefaultSubagentModelaccept it.
Prices and the (N×) labels always stay the plugin's, because the CLI carries none. A model the installed CLI does not know at all (an old Claude Code and a new model) keeps the plugin's numbers, and /claude-code-doctor names it with the release that fixes it. That includes a model older than the Claude Code release that added it even when the CLI answers with a name: Claude Code 2.1.314 calls claude-fable-5-2 "Fable" at its 200k fallback window, because it files every Fable id under its picker's Fable entry, so the plugin goes by the release in Claude Code's changelog (2.1.321) instead.
How it is read, and what it costs. Once per Claude Code version, in the background, the plugin starts one short claude --print --bare --model <id> per registered model (18 now, about 3 seconds in all on Claude Code 2.1.314) and asks each two read-only questions over the control channel (initialize and get_context_usage), never a message. --bare reads no login, and the plugin removes ANTHROPIC_API_KEY from the probe's environment, so there is nothing a request could be billed to; measured against a local server standing in for the API, the probe made no request at all. The answer is kept in $XDG_STATE_HOME/opencode-claude-code-plugin/cli-models.json (0600), keyed on the real claude binary and its version, so it is read again only after claude update.
Startup never waits for it. On opencode 2 a new answer reloads the providers in the same session; on opencode 1 it applies from the next start. If the probe is killed by its own deadline (a very busy machine) nothing is stored and the next start asks again; if the CLI refuses it (too old for these requests), that is remembered for a day. Either way the plugin's own list is registered. /claude-code-doctor has a Model catalog from the Claude CLI section that says which version answered and what it changed. "cliModelCatalog": false turns all of it off.
Variants set the underlying reasoning effort. They're regular opencode model variants, so pick them in the model selector. If you'd previously declared variants in your project's opencode.json, they're merged on top of the defaults so nothing gets lost.
To rename a model, change a limit, or add a custom one:
{
"plugin": ["@khalilgharbaoui/opencode-claude-code-plugin"],
"provider": {
"claude-code": {
"models": {
"claude-sonnet-4-6": {
"name": "Sonnet (custom)",
"limit": { "context": 200000, "output": 32768 }
}
}
}
}
}Anything you supply is merged on top of the defaults; you don't need to redeclare every model.