Grok 4.3
Grok 4.3 is a chat model in xAI's Grok series, measured at roughly 112 output tokens per second with about 0.96s to first token in Artificial Analysis testing.
API Pricing
Cheapest on Velokey — 50% below avg| Provider | Input / 1M | Output / 1M | Cached / 1M |
|---|---|---|---|
| $0.500 | $1.00 | $0.080 | |
| $1.00 | $2.00 | $0.160 | |
| $1.25 | $2.50 | $0.200 | |
| $1.25 | $2.50 | $0.200 |
Prices updated daily. Last check: Oct 11, 2026
Compare API pricing for every Grok model →Grok 4.3 pricing by provider
Input, output, and batch rates, plus alternatives, for one provider at a time.
Performance & Benchmarks
Source: Artificial Analysis →Reasoning & Knowledge
- GPQA Diamond90.1%
- Humanity's Last Exam37.2%
Coding
- SciCode48.3%
Agentic & Tool Use
- Terminal-Bench Hard37.9%
- Terminal-Bench v2.139.7%
- τ²-bench97.7%
- τ-bench Banking12.4%
Instruction & Long Context
- IFBench81.3%
- Long-Context Reasoning73.0%
Benchmarks measured Oct 2026. Scores are independent evaluations, not vendor-reported.
Model Details
General
- Creator
- SpaceXAI
- Modalities
- Text
Capabilities
- Open Source
- No
Strengths & Limitations
Strengths
- Measured at approximately 112 output tokens per second in Artificial Analysis testing, a rate suited to streaming chat output
- Time to first token of roughly 961 ms in the same testing, keeping initial response latency under one second
- Part of the Grok 4.x generation, so it inherits the API shape and integration patterns already used by earlier Grok 4 deployments
- Available through an API endpoint, allowing drop-in use in existing chat-completion client code
- Latency and throughput figures come from independent third-party measurement rather than vendor-published claims
- Listed alongside competing models in the pricing table on this page for direct provider-by-provider cost comparison
Limitations
- We do not track a confirmed context window size for Grok 4.3, so long-document workloads need verification against xAI's documentation
- Supported input modalities are not confirmed in our metadata — do not assume image or audio input without checking
- No verified benchmark scores for coding, math, or reasoning are available in our database for this model
- Provider availability for Grok models is narrower than for widely-hosted open-weight alternatives, which limits routing and failover options
- Reported throughput and time-to-first-token figures reflect one measurement setup and may differ under production load
Key Features
About Grok 4.3
Common Use Cases
Grok 4.3's measured profile — sub-second time to first token combined with a triple-digit token generation rate — makes it a reasonable candidate for interactive workloads where users watch text appear: assistant-style chat interfaces, customer-facing support bots, in-product copilots, and drafting or rewriting tools. The same characteristics suit multi-turn agent loops where several sequential model calls compound latency, since first-token delay is paid on every hop. For batch workloads such as bulk summarization or offline classification, latency matters less and the deciding factors become per-token cost and accuracy on your specific task — compare the pricing table on this page and run your own evaluation set. Because we do not have a confirmed context window or modality list for Grok 4.3, verify those specs directly with xAI before committing it to long-context document processing or any pipeline that depends on non-text input.
Frequently Asked Questions
How much does Grok 4.3 cost to use?
Pricing varies by provider and by pricing type — input tokens, output tokens, and any cached or batch rates are typically billed separately, and providers change rates over time. Check the pricing table on this page for the current figures across the providers we track.
What is Grok 4.3 best used for?
Its measured latency profile — roughly 961 ms to first token and about 112 output tokens per second — fits interactive uses such as chat assistants, in-product copilots, drafting tools, and agent loops where several sequential calls add up. For offline batch work, cost per token and task accuracy usually matter more than response speed.
How fast is Grok 4.3?
Artificial Analysis measured Grok 4.3 at approximately 111.99 output tokens per second with a time to first token of about 961 milliseconds. Those numbers reflect a specific test setup; actual speed depends on the provider, prompt length, region, and current load.
Does Grok 4.3 accept image input?
Our database does not record confirmed modality information for Grok 4.3, so we cannot state which input types it supports. Check xAI's model documentation for the authoritative list before building a pipeline that depends on non-text input.
How does Grok 4.3 relate to other Grok models?
It belongs to the Grok 4.x generation from xAI, making it an iteration within that line rather than a separate small or specialized variant. For a like-for-like comparison against other Grok entries or models from other labs, use the pricing table and the measured latency figures on this page, then validate on your own evaluation set.