grok-4.7

Tool-calling and vision together support browser automation agents, though family throughput (57.8 tok/s on grok-4.6) is a latency concern for interactive loops.

Vision capability supports chart data extraction workflows, though unverified against vision-specific benchmarks.

Extended-thinking mode supports multi-step analytical reasoning, though no data-analysis-specific benchmarks are published for this variant.

Vision capability supports document OCR workflows, though no OCR-specific benchmark is documented for this variant.

Vision capability enables image Q&A, but no vision benchmark scores are published for this specific variant.

Large context window supports long-document generation, though no family-specific writing benchmarks are documented for this variant.

Family throughput of 57.8 tokens/sec on grok-4.6 makes this a poor fit for latency-sensitive, real-time chatbot interactions.

No speech capability documented and family-level low throughput (57.8 tok/s) make this unsuitable for real-time voice agent latency requirements.