Compare models

Put up to 4 models beside each other — token prices, context windows, capabilities and provider, from the same catalogue the model pages read.

Grok STT 1.0xAI Remove
Nova-3Deepgram Remove
Qwen3 ASR 1.7BQwen Remove
Grok 4.7SpaceXAI Remove

4 is the maximum. Remove one to add another.

grok-stt-1.0 vs nova-3 vs qwen3-asr-1.7b vs grok-4.7
Attribute	Grok STT 1.0grok-stt-1.0	Nova-3nova-3	Qwen3 ASR 1.7Bqwen3-asr-1.7b	Grok 4.7grok-4.7
Pricing
Input	$0 / 1M	$0 / 1M	$0 / 1M	$1.60 / 1M
Output	$0 / 1M	$0 / 1M	$0 / 1M	$4.80 / 1M
Cache Write (5m)	Not applicable	Not applicable	Not applicable	$1.60 / 1M
Cache Write (1h)	Not applicable	Not applicable	Not applicable	$1.60 / 1M
Cache Read	Not applicable	Not applicable	Not applicable	$1.60 / 1M
Web Search	$0 / 1M	$0 / 1M	$0 / 1M	$0 / 1M
Context
Max context	N/A	N/A	N/A	500K
Max output	N/A	N/A	N/A	N/A
Capabilities
Vision	No	No	No	Yes
Function Calling	No	No	No	Yes
JSON Mode	No	No	No	Yes
Streaming	No	No	No	Yes
Catalogue
Provider	xAI	Deepgram	Qwen	SpaceXAI
Category	voice	voice	voice	chat
Charge type	Pay As You Go	Pay As You Go	Pay As You Go	Pay As You Go
Released	2026-07-23	2026-07-15	2026-08-13	—
Description
Summary	xAI Grok STT 1.0 speech-to-text. Billed per second of audio.	Deepgram Nova-3 speech-to-text. Billed per second of audio.	Qwen3-ASR 1.7B speech-to-text. Billed per second of audio.	Grok 4.7 is SpaceXAI's flagship model for coding, agentic workflows, and professional knowledge work, succeeding Grok 4.6. It is particularly strong at long-running software engineering, self-verification, and long-context execution, while improving capabilities in document drafting, presentations, and other professional tasks. Trained with extended reinforcement learning focused on multi-hour problems, Grok 4.7 is optimized for sustained, complex task execution and natively supports the Grok Bot harness for conversational workflows. It also introduces an enhanced safeguard stack designed to combine strong jailbreak resistance with low refusal rates for legitimate technical work. Reported benchmark results use xhigh reasoning effort.