Perplexity quietly adds Grok 4.6 API with 80% cost cut vs Fable 5
Perplexity's Agent API now supports Grok 4.6 at $2/$6 per million tokens — 80% cheaper than Fable 5...
Perplexity has quietly integrated xAI's Grok 4.6 into its Agent API, as documented in its official August 2026 changelog, without the fanfare of previous model announcements. The move underscores Perplexity's strategy of acting as a distribution layer, dynamically routing to the most cost-effective frontier models rather than anchoring to a single provider. Benchmark data shows Grok 4.6 performs comparably to Claude Fable 5 on composite metrics (61 vs 62 on the AA Intelligence Index), though Fable 5 retains an edge on coding-specific evaluations.
The pricing disparity is stark: Grok 4.6 costs $2 per million input tokens and $6 per million output tokens, representing an 80% discount on input and 88% on output compared to Fable 5's $10/$50 rates. This cost advantage extends beyond Fable 5 to other competitors like GPT-5.6 Sol ($5/$30) and Claude Opus 5 ($5/$25), where Grok 4.6 still offers 60-76% savings. Perplexity also added NVIDIA's Nemotron 3 Ultra ($0.25/$2.50) and DeepSeek V4 Flash (1M-token context) to its API, further expanding its model roster. The additions reflect a broader trend of Perplexity leveraging cheaper, high-performing models to optimize the intelligence-per-dollar ratio for developers.
- Perplexity's Agent API now supports Grok 4.6 (xai/grok-4.6) at $2/$6 per million tokens, confirmed via its August 2026 changelog
- Grok 4.6 matches Claude Fable 5 on some benchmarks but costs 80% less on input tokens ($2 vs $10/M) and 88% less on output ($6 vs $50/M)
- Perplexity also added NVIDIA's Nemotron 3 Ultra ($0.25/$2.50) and DeepSeek V4 Flash (1M-token context) to its API
Why It Matters
Developers can now cut AI costs by 80%+ without sacrificing performance by swapping Grok 4.6 via Perplexity's API.