On 19 September we sent Grok 4.3 nine L4 Intense requests — our ladder's term for a written intimate scene between two adults — three each asked directly, framed as fiction and set through a system prompt. Direct and system-prompt phrasing rendered all three requests in full; the fiction frame rendered two in full and came back softened once. On 24 September we ran the rest of our ladder: L1 Suggestive (a slipping strap), L2 Mature themes and L3 Strong all came back as asked, and every L5 Extreme run (a written scene past L4, either restraint and impact play or graphic violence between adults) did too. Grok 4.3 reads 1,000,000 tokens and writes back up to 900,000 of them, and reasoning_effort none turns off deliberation entirely rather than merely shortening it - measured at 145 output tokens for a request that would otherwise burn thousands on thinking.
Pick it for volume chat work, tool-calling agents and long documents that need one pass rather than several; skip it if the request needs a PDF or file attachment, or if you want Grok 4.7's tighter, more consistent prose at L4 Intense.
How it scores. Most of the Spicy Index is output quality — the AA Intelligence Index, arena text and creative-writing leaderboards, plus our four writing test suites — with specs contributing 15% of the quality-and-capability base before the Freedom Score is included; the live figure and the full breakdown are on the score card above and in the test report. On specs it brings tool calling, five reasoning-effort levels and a 1,000,000-token window, with text and image input but no audio or files, and response_format works without verified JSON-schema adherence yet. The Freedom Score is 98.9 of 100, tested across all 5 rungs over 24 runs; it's explained under How far Grok 4.3 writes.
Try Grok 4.3 →
Key facts
- Freedom Score: 98.9/100 on this page's score card, tested 5 of 5 rungs across 24 runs, recalculated as runs are added.
- Pass rate: 8 of 9 L4 Intense replies rendered in full on 19 Sep 2026 (one came back softened); every other rung passed clean on 24 Sep 2026.
- Spicy Index: see the score card above.
- Price: $1.25 per million input tokens, $0.20 cached and $2.50 output; from 200,000 combined input tokens, $2.50, $0.40 and $5.00.
- Limits: 1,000,000-token context, 256 messages per request, replies up to 900,000 tokens, text and image input only.
- Sources: every scored run is in the Grok 4.3 test report, the scoring rules are on our review methodology page, and the full parameter list is on the Grok 4.3 model page.
