The differences that matter
Put any two—or all three—side by side. Your selection stays in the URL so you can bookmark or share it.
Standard API prices in USD per million tokens. Verification dates appear below; rates follow the published schedule. Conditions appear alongside the rates. Task guidance follows documented capabilities; it is not a performance ranking.
| What matters | Claude Fable 5.1 | GPT-6 Astra | Gemini 4 Argon |
|---|---|---|---|
| Consider it for | Long-running agents | Reasoning & coding | Audio, video & long-form work |
| Why it may fit | Anthropic positions this model for demanding reasoning and long-horizon agentic work. Adjustable effort lets you test the tradeoff between depth, time and token use. | Documented reasoning controls (low, medium, high, xhigh, max) let you tune effort. Check results and cost on your own tasks. | The API accepts audio and video directly, alongside text and images. This can avoid separate transcription and frame-extraction steps. |
| Tradeoff | A larger context budget does not establish better recall or answer quality. Audio/video workflows need conversion unless documented separately by the provider. | A larger context budget does not establish better recall or answer quality. Long prompts increase the rate for the whole request. Audio/video workflows need conversion unless documented separately by the provider. | Input and output consume the same token budget. Long decode is a separate mode; the default output limit is lower. |
| Inputs | Text, images, PDF | Text, images, PDF | Text, images, video, audio, PDF |
| Context budget | 1,000,000 tokens | 1,050,000 tokens · total context; up to 922,000 tokens input | 1,048,576 tokens · shared input + output |
| Maximum output | 128,000 tokens | 128,000 tokens | 65,536 tokens default; up to 1,048,576 tokens with long decode |
| Reasoning controls | low, medium, high, xhigh, max; default high; Adaptive (always on) | low, medium, high, xhigh, max; Reasoning supported | low, medium, high |
| Input / 1M tokens | $10 | $10 | $2 |
| Output / 1M tokens | $50 | $50 | $10 |
| Pricing conditions | Standard Claude API token rates. Caching, batch processing, tools and cloud-platform rates are separate. | Above 272,000 input tokens, the full request costs 2× input and 1.5× output. Standard text-token rates. Caching, Batch, Flex, Fast mode and tools have separate rates. | Standard paid API rates; output includes thinking tokens. Caching, batch, service tiers and grounding have separate rates. Every processed token is billed, including internal tool system instructions and function declarations when grounding. Google Search grounding is billed per individual search query. From 2026-12-01: $4 input / $20 output per million tokens (UTC). |
| API access | Claude API; cloud-platform availability and prices can differ | OpenAI API | Paid Gemini API; unavailable on the free tier |
| Native output | text | text | text |
| Official sources | Current model lineupModel specificationsReasoning controlsPDF support | Current model lineupModel specificationsPDF support | Current model lineupModel specificationsAPI pricing |
Check the verification dates
Claude Fable 5.1 ·
Official sources checked automatically every two hours.
GPT-6 Astra ·
Official sources checked automatically every two hours.
Gemini 4 Argon ·
The latest refresh could not be accepted. Showing the last verified profile.
Task guidance follows documented capabilities. No cross-lab performance ranking. API limits and prices differ from chat subscriptions. Open a profile for the official sources and pricing conditions.