xAI model product
Grok 4.5
5 is SpaceXAI's frontier model for coding, agentic tasks, and knowledge work.
Updated Aug 12, 2026. Default version: Grok 4.5
LLMBoard score82.2Grok 4.5
Coverage40%9 benchmark families
Context window500KTokens
Official input price$2xAI API
Grok 4.5 Specifications
Technical details for the model's default version.
- Version
- Grok 4.5
- Released
- Jul 16, 2026
- Knowledge cutoff
- Feb 1, 2026
- Parameters
- N/A
- Context window
- 500K
- Max output
- 500K
- Inputs
- image, text
- Outputs
- text
- Open weights
- No
- License
- Proprietary
Grok 4.5 Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Grok 4.5 Benchmark Results
Benchmark scores for Grok 4.5.
| AA-Omniscience Index | 126 points | 01 | 1 | 100.0% | B | |
| AutomationBench-AA | 51.4% | 01 | 1 | 100.0% | B | |
| DeepSWE 1.0 | 62.0% | 01 | 1 | 100.0% | C | |
| OmniScience (non-hallucination rate) | 46.0% | 01 | 1 | 100.0% | B | |
| Tau3 Banking | 33.0% | 01 | 3 | 100.0% | B | |
| OmniScience | 52.0% | 02 | 2 | 0.0% | B | |
| SWE-Marathon | 29.0% | 02 | 3 | 50.0% | C | |
| Artificial Analysis | 54.0% | 03 | 5 | 50.0% | B | |
| SWE-Bench Pro | 64.7% | 05 | 45 | 90.9% | C | |
| DeepSWE | 53.0% | 06 | 10 | 44.4% | C | |
| FrontierCode 1.1 | 42.4% | 07 | 15 | 57.1% | B | |
| Terminal-Bench 2.1 | 83.3% | 07 | 19 | 66.7% | C | |
| GPQA | 93.0% | 09 | 234 | 96.6% | B | |
| DeepSWE 1.1 | 54.0% | 12 | 20 | 42.1% | B |
Grok 4.5 Arena Results
Preference and agent-evaluation results for the default version.
| agent bash recovery steps | overall | 07 | 0.1 | N/A | 23.4K | |
| agent tool hallucination | overall | 08 | 0.0 | N/A | 1.2M | |
| webdev | overall | 12 | 1553.8 | 5,380 | N/A | |
| vision style control | overall | 13 | 1285.0 | 3,871 | N/A | |
| agent | overall | 15 | 0.1 | N/A | 1.3M | |
| agent task outcome explicit | overall | 15 | 0.1 | N/A | 18.7K | |
| agent praise complaint | overall | 17 | 0.1 | N/A | 5.9K | |
| vision | overall | 17 | 1295.4 | 3,871 | N/A | |
| agent steerability | overall | 18 | 0.0 | N/A | 24K | |
| document | overall | 20 | 1453.7 | 2,435 | N/A | |
| document style control | overall | 21 | 1462.0 | 2,435 | N/A | |
| text factuality | overall | 23 | 1467.3 | 17,125 | N/A | |
| text style control | overall | 34 | 1469.4 | 17,149 | N/A | |
| text | overall | 36 | 1451.7 | 17,149 | N/A |
Grok 4.5 Runtime Performance
Provider-specific output speed and catalog latency for Grok 4.5. Runtime does not affect the capability score.
| xAI | 6.526 tok/s | 1.73 s | 500K | N/A |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Grok 4.5 Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- $2 input, $6 output per 1M
- Official provider
- xAI
- Lowest third-party
- From $2 input, $6 output per 1M via Impossibl
- Tracked offerings
- 20
| Kenari | grok-4-5 | global | N/A | N/A | 500K | |
| Impossibl | xai/grok-4.5 | global | $2 | $6 | 500K | |
| CrossModel | x-ai/grok-4.5 | global | $2 | $6 | 500K | |
| OpenCode Go | grok-4.5 | global | $2 | $6 | 500K | |
| NanoGPT | x-ai/grok-4.5 | global | $2 | $6 | 500K | |
| DaoXE | grok-4.5 | global | $2 | $6 | 500K | |
| LLM Gateway | grok-4-5 | global | $2 | $6 | 500K | |
| OpenCode Zen | grok-4.5 | global | $2 | $6 | 500K | |
| GitHub Copilot | grok-4.5 | global | $2 | $6 | 500K | |
| Pioneer | grok-4.5 | global | $2 | $6 | 500K | |
| ZenMux | x-ai/grok-4.5 | global | $2 | $6 | 500K | |
| Abacus | grok-4.5 | global | $2 | $6 | 500K | |
| xAI | grok-4.5 | global | $2 | $6 | 500K | |
| AIHubMix | grok-4.5 | global | $2 | $6 | 1M | |
| Requesty | grok-4.5 | global | $2 | $6 | 500K | |
| Kilo Gateway | x-ai/grok-4.5 | global | $2 | $6 | 500K | |
| Merge Gateway | xai/grok-4.5 | global | $2 | $6 | 500K | |
| Vercel AI Gateway | xai/grok-4.5 | global | $2 | $6 | 500K | |
| OpenRouter | x-ai/grok-4.5 | global | $2 | $6 | 500K | |
| Venice AI | grok-4-5 | global | $2.27 | $6.8 | 500K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Grok 4.5 Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
| Grok 4.5 | 82.2 | N/A | 500K | 500K | No | Proprietary |
What is Grok 4.5?
Key information about Grok 4.5 and its available data.
5 is SpaceXAI's frontier model for coding, agentic tasks, and knowledge work. Trained alongside Cursor across tens of thousands of NVIDIA GB300 GPUs, it uses large-scale reinforcement learning focused on multi-step software engineering and other technical work.
The model supports text and image inputs, reasoning, function calling, structured outputs, and a 500K-token context window, and is served at 80 tokens per second.
Data as of 2026-08-11.
Grok 4.5 vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to Grok 4.5
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
XA
Grok 4 Heavy
xAI
XA
Grok 4.20 Beta Reasoning
xAI
XA
Grok 4
xAI
XA
Grok 3
xAI
XA
Grok 4 Fast
xAI
XA
Grok 4.1 Thinking
xAI
FAQ
Common questions about Grok 4.5.
When was Grok 4.5 released?
Grok 4.5's default version was released on Jul 16, 2026.
How much does Grok 4.5 cost?
Grok 4.5's official API price is $2 per million input tokens and $6 per million output tokens via xAI. The lowest tracked third-party offer starts at $2 input and $6 output via Impossibl.
Who created Grok 4.5?
Grok 4.5 was created by xAI.
What is the context window for Grok 4.5?
The default version has a 500K token context window.
Is Grok 4.5 open weight?
No. The default version is not marked as having publicly available weights.
How many API providers offer Grok 4.5?
20 provider offerings are linked to the default version.
What models should I compare Grok 4.5 with?
Nearby ranked alternatives include Claude Sonnet 5, GPT-5.5, Muse Spark 1.2.