Model cardPro
One model at a glance: inputs, outputs, context and prices swap in place when you change model, a volume slider turns prices into a monthly bill, benchmark rows rank it, and Compare sets a rival beside it and ticks the stronger figure.
oxyn add model-cardClaude Sonnet 4.5
AnthropicThe balanced Claude for everyday coding, agents and analysis at scale.
- Text
- Image
- Audio
- Video
- Tools
- Reasoning
- Context window
- 1M tokens
- Max output
- 64K tokens
- Knowledge cutoff
- Jul 2025
- Input per 1M
- $3.00
- Output per 1M
- $15.00
- Cached per 1M
- $0.30
Benchmarks
- Claude Sonnet 4.577.2%
- Claude Opus 4.174.5%
- GPT-574.9%
SWE-bench Verified, real GitHub issues resolved. Sample prices and scores, not official.
| Spec | Claude Sonnet 4.5 | GPT-5 |
|---|---|---|
| Modalities | Text, Image, PDF, Tools, Reasoning | Text, Image, PDF, Audio, Tools, Reasoning |
| Context window | 1M tokens, better by 2.5x | 400K tokens |
| Max output | 64K tokens | 128K tokens, better by 2.0x |
| Knowledge cutoff | Jul 2025, better by +10 mo | Sep 2024 |
| Input price | $3.00 | $1.25, better by −58% |
| Output price | $15.00 | $10.00, better by −33% |
| Cost at 100M/mo | $600 | $344, better by −43% |
| SWE-bench | 77.2%, better by +2.3 | 74.9% |
| GPQA | 83.4% | 85.7%, better by +2.3 |
| Terminal-Bench | 50.0%, better by +6.2 | 43.8% |
GPT-5 leads on 5 of 10, Claude Sonnet 4.5 on 4. Prices per 1M tokens; cost assumes 75% input. Sample figures, not official.
claude-sonnet-4-5The preview stays free for everyone. Pro opens the code of every Pro component, screen, agent-app template, block and kit, plus the Pro CLI.
Rendered from Rust at build time · goes live when visible
When to use
- Decide whether a model fits a job: what it reads and writes, its context and prices, an estimated monthly cost at your volume, and how it scores against rivals.
When not to use
- Replays a multi-agent run on one clock: use Agent timeline instead.
- A small automation on a single rail: use AI workflow instead.
- Four variations grow out of a grid of dots: use Image generation instead.
Installation
Model card is part of Oxyn Pro; the preview is free. Pro items install with the oxyn CLI once your licence is signed in; until then Pro members copy the code from this page.
oxyn add model-cardUsage
The preview is free. Pro opens the code of every Pro component, screen and template.
Examples
Details
Details of Claude Sonnet 4.5 at 100M tokens a month.
Claude Sonnet 4.5
AnthropicThe balanced Claude for everyday coding, agents and analysis at scale.
- Text
- Image
- Audio
- Video
- Tools
- Reasoning
- Context window
- 1M tokens
- Max output
- 64K tokens
- Knowledge cutoff
- Jul 2025
- Input per 1M
- $3.00
- Output per 1M
- $15.00
- Cached per 1M
- $0.30
Benchmarks
- Claude Sonnet 4.577.2%
- Claude Opus 4.174.5%
- GPT-574.9%
SWE-bench Verified, real GitHub issues resolved. Sample prices and scores, not official.
| Spec | Claude Sonnet 4.5 | GPT-5 |
|---|---|---|
| Modalities | Text, Image, PDF, Tools, Reasoning | Text, Image, PDF, Audio, Tools, Reasoning |
| Context window | 1M tokens, better by 2.5x | 400K tokens |
| Max output | 64K tokens | 128K tokens, better by 2.0x |
| Knowledge cutoff | Jul 2025, better by +10 mo | Sep 2024 |
| Input price | $3.00 | $1.25, better by −58% |
| Output price | $15.00 | $10.00, better by −33% |
| Cost at 100M/mo | $600 | $344, better by −43% |
| SWE-bench | 77.2%, better by +2.3 | 74.9% |
| GPQA | 83.4% | 85.7%, better by +2.3 |
| Terminal-Bench | 50.0%, better by +6.2 | 43.8% |
GPT-5 leads on 5 of 10, Claude Sonnet 4.5 on 4. Prices per 1M tokens; cost assumes 75% input. Sample figures, not official.
claude-sonnet-4-5Code Pro
The preview is free. Pro opens the code of every Pro component, screen and template.
Compare
GPT-5 against a rival with the better values checked.
GPT-5
OpenAIOpenAI's flagship for reasoning, coding and multimodal work.
- Text
- Image
- Audio
- Video
- Tools
- Reasoning
- Context window
- 400K tokens
- Max output
- 128K tokens
- Knowledge cutoff
- Sep 2024
- Input per 1M
- $1.25
- Output per 1M
- $10.00
- Cached per 1M
- $0.12
Benchmarks
- GPT-574.9%
- Claude Opus 4.174.5%
- Claude Sonnet 4.577.2%
SWE-bench Verified, real GitHub issues resolved. Sample prices and scores, not official.
| Spec | GPT-5 | Gemini 2.5 Pro |
|---|---|---|
| Modalities | Text, Image, PDF, Audio, Tools, Reasoning | Text, Image, PDF, Audio, Video, Tools, Reasoning |
| Context window | 400K tokens | 1M tokens, better by 2.5x |
| Max output | 128K tokens, better by 2.0x | 64K tokens |
| Knowledge cutoff | Sep 2024 | Jan 2025, better by +4 mo |
| Input price | $1.25 | $1.25 |
| Output price | $10.00 | $10.00 |
| Cost at 100M/mo | $344 | $344 |
| SWE-bench | 74.9%, better by +7.7 | 67.2% |
| GPQA | 85.7% | 86.4%, better by +0.7 |
| Terminal-Bench | 43.8%, better by +18.5 | 25.3% |
GPT-5 leads on 3 of 10, Gemini 2.5 Pro on 3. Prices per 1M tokens; cost assumes 75% input. Sample figures, not official.
gpt-5Code Pro
The preview is free. Pro opens the code of every Pro component, screen and template.
Picker pane
Shown as the side panel of a model chooser, with a Use model button (Oxyn extra).
Gemini 2.5 Pro
GoogleGoogle's thinking model with a long context and native audio and video.
- Text
- Image
- Audio
- Video
- Tools
- Reasoning
- Context window
- 1M tokens
- Max output
- 64K tokens
- Knowledge cutoff
- Jan 2025
- Input per 1M
- $1.25
- Output per 1M
- $10.00
- Cached per 1M
- $0.31
Benchmarks
- Gemini 2.5 Pro67.2%
- Claude Opus 4.174.5%
- Claude Sonnet 4.577.2%
SWE-bench Verified, real GitHub issues resolved. Sample prices and scores, not official.
| Spec | Gemini 2.5 Pro | GPT-5 |
|---|---|---|
| Modalities | Text, Image, PDF, Audio, Video, Tools, Reasoning | Text, Image, PDF, Audio, Tools, Reasoning |
| Context window | 1M tokens, better by 2.5x | 400K tokens |
| Max output | 64K tokens | 128K tokens, better by 2.0x |
| Knowledge cutoff | Jan 2025, better by +4 mo | Sep 2024 |
| Input price | $1.25 | $1.25 |
| Output price | $10.00 | $10.00 |
| Cost at 100M/mo | $344 | $344 |
| SWE-bench | 67.2% | 74.9%, better by +7.7 |
| GPQA | 86.4%, better by +0.7 | 85.7% |
| Terminal-Bench | 25.3% | 43.8%, better by +18.5 |
Gemini 2.5 Pro leads on 3 of 10, GPT-5 on 3. Prices per 1M tokens; cost assumes 75% input. Sample figures, not official.
gemini-2-5-proCode Pro
The preview is free. Pro opens the code of every Pro component, screen and template.
Error
Couldn't load, with Retry.
Claude Sonnet 4.5
AnthropicThe balanced Claude for everyday coding, agents and analysis at scale.
- Text
- Image
- Audio
- Video
- Tools
- Reasoning
- Context window
- 1M tokens
- Max output
- 64K tokens
- Knowledge cutoff
- Jul 2025
- Input per 1M
- $3.00
- Output per 1M
- $15.00
- Cached per 1M
- $0.30
Benchmarks
- Claude Sonnet 4.577.2%
- Claude Opus 4.174.5%
- GPT-574.9%
SWE-bench Verified, real GitHub issues resolved. Sample prices and scores, not official.
| Spec | Claude Sonnet 4.5 | GPT-5 |
|---|---|---|
| Modalities | Text, Image, PDF, Tools, Reasoning | Text, Image, PDF, Audio, Tools, Reasoning |
| Context window | 1M tokens, better by 2.5x | 400K tokens |
| Max output | 64K tokens | 128K tokens, better by 2.0x |
| Knowledge cutoff | Jul 2025, better by +10 mo | Sep 2024 |
| Input price | $3.00 | $1.25, better by −58% |
| Output price | $15.00 | $10.00, better by −33% |
| Cost at 100M/mo | $600 | $344, better by −43% |
| SWE-bench | 77.2%, better by +2.3 | 74.9% |
| GPQA | 83.4% | 85.7%, better by +2.3 |
| Terminal-Bench | 50.0%, better by +6.2 | 43.8% |
GPT-5 leads on 5 of 10, Claude Sonnet 4.5 on 4. Prices per 1M tokens; cost assumes 75% input. Sample figures, not official.
claude-sonnet-4-5Code Pro
The preview is free. Pro opens the code of every Pro component, screen and template.
API reference
ModelCard
| Prop | Type | Default | Description |
|---|---|---|---|
nowrequired | ClockMs | — | Host clock. |
model | usize | 1 | Starting model (0 Opus 4.1, 1 Sonnet 4.5, 2 GPT-5, 3 Gemini 2.5 Pro, 4 Grok 4). |
view | String | "details" | `details` or `compare`. |
volume | usize | 5 | Starting volume stop (0–8: 1M … 1B tokens a month). |
loading | bool | Default::default() | Show the skeleton. |
error | bool | Default::default() | Show the error state. |
selectable | bool | Default::default() | Show the Use model button (model picker detail pane). |
tabs | bool | true | Show the model tabs above the card. |
Read from the Rust signature and doc comments.
Keyboard interactions
| Key | Action |
|---|---|
| Tab | Model tabs, Compare, the volume slider, benchmark switcher and rows, then Copy and Use model |
| ←or→ | Model tabs and the benchmark switcher |
| Arrows · Page upordown · HomeorEnd | Step the monthly volume |
| Enteror↓ | Open the rival menu; ↑/↓ move, Enter picks, Escape closes |
| SpaceorEnter | Toggle Compare, copy the id, select the model |
Accessibility
- The card is an article labelled by the model; specs and prices are description lists.
- Compare is a real table with headers on both axes; the stronger figure gets a tick plus the words better by, not just a colour.
- Benchmark rows are focusable and read the full readout (score, rank, gap to the leader); the bars are decorative.
- A modality the model lacks gets a dashed outline, a dimmer tone and the words not supported.
- Loading sets aria-busy; errors use role alert; the polite region says what changed.
From the spec
Tabs are a tablist with arrow keys; the volume slider is a native range with a text value; compare marks the better value with text (Better) as well as the check. Every action returns the line the polite live region announces. The focus outline shows only for keyboard focus.
Motion & interaction
Switching model rolls every number in place; the card never changes size because every panel shares one grid cell. The volume slider moves the monthly estimate live. Copy model id flashes Copied for 1.2 s.
Timing
Everything on screen is derived from the host clock now (no timers in the component), so playback, tests and the GPUI view agree.
| Transition | Duration | Easing |
|---|---|---|
| Value roll on model switch | 180 ms | ease-out |
| Panel swap (tabs, compare) | 160 ms crossfade | ease-out |
| Copied flash | 1200 ms | step |
Reduced motion: streamed text and moves appear at their end state; progress still advances, without travel animation.
Responsive behaviour
- All four panels (details, compare, loading, error) stack in the same cell, so switching never resizes the card.
- Below 560 px the spec tiles turn into rows, the monthly figure drops below the slider and benchmark bars sit beneath their names.
- The compare table scrolls inside its own frame rather than the page.
Edge cases (spec)
Long model names wrap to two lines in the header; the tab row scrolls horizontally under 390 px instead of wrapping onto prices. Prices use tabular figures. Compare collapses to one column of pairs under 640 px. Checked at 1440, 768 and 390 px in light and dark with no text overlap. Chrome is monochrome grey; colour is semantic only (success, warning, danger, info) plus the providers' coloured brand marks.
Notes for AI
- Import
ModelCardfromoxyn_dioxusand includeSTATUS_INDICATOR_CSSplus the token stylesheet once. - A GPUI version exists:
AiKitViewinoxyn_gpui. - Props to set first:
now. - It reads a clock (
now) for live timers; pass one clock for the whole page rather than one per component. - Keep its states and timings as they ship; restyle it through the CSS variables rather than by hand.