Model
Model explorer

Grok 4.5

CLOSED
SpaceXAI · Grok 4.5 family · released Jul 8, 2026

SpaceXAI's newest flagship (announced July 8, 2026, public July 9), the first model shipped since the SpaceXAI rebrand and since acquiring Cursor; positioned by Elon Musk as 'Opus-class... but faster, more token-efficient and lower cost' and 'roughly comparable to Opus 4.7, but much faster.' Powers the Grok Build CLI, is available in Cursor on all plans, and via the SpaceXAI console; not yet available in the EU at launch (targeted mid-July 2026). Ranks 4th on the Artificial Analysis Intelligence Index (score 54) behind Fable 5, GPT-5.5, and Opus 4.8, but at roughly 60%+ lower price than Opus 4.8/GPT-5.5, with markedly lower average output-token usage per task (token efficiency claim).

ReasoningCodingVisionFunction callingTool useAgentic
2444.0
Elo · rank #38
Parameters
Undisclosed
Active params
Undisclosed
Context
500K tokens
Architecture
Mixture-of-Experts (size undisclosed by SpaceXAI); internally referred to in press coverage as a 'V9' generation, a full architecture shift from the V8-series behind Grok 4.3, retrained heavily on real Cursor developer-session data for coding/agentic work
License
Proprietary
Languages
API price (in/out)
$2 / $6
Modalities
text · vision
Benchmark results
Bar shows position within the tracked field; marker = field best
Arena EloHuman preference1464#5
best: Claude Fable 5 · 1505
CursorBenchCoding66.7%#6
best: Claude Fable 5.1 · 73.4%
FrontierBenchAgents17.8%#6
best: Claude Opus 5 · 43.3%
best: Claude Fable 5 · 64.9%
GDPval-AAAgents1539#19
best: Claude Fable 5 · 1932
GPQA DiamondReasoning93.1%#12
best: GPT-6 Astra · 96.0%
best: Grok 4.6 · 15.8%
Humanity's Last ExamReasoning40.3%#28
best: Claude Opus 5 · 64.7%
LiveCodeBenchCoding87.3%#17
best: DeepSeek-V4-Pro (Think Max) · 93.5%
MMLU-ProKnowledge89.2%#4
best: Claude Fable 5 · 91.5%
MMMU-ProVision80.4%#12
best: Claude Opus 4.7 · 85.5%
MMMUVision61.8%#75
best: Claude Fable 5 · 89.3%
SWE-bench ProCoding64.7%#6
best: Claude Fable 5.1 · 81.2%
SWE-bench VerifiedCoding86.6%#5
best: Claude Opus 5 · 96.0%
τ³-BankingAgents33.0%#2
best: GPT-5.6 Sol · 33.0%
Terminal-Bench 2.0Coding83.3%#13
best: Gemini 3.8 Flash · 89.4%
Terminal-Bench 3.0Agents15.7%#7
best: GPT-5.6 Sol · 34.6%
Run it locally
Closed weights — available via API only. No local deployment.
Input / M tok
$2
Output / M tok
$6
Grok 4.5 family
Elo progression across releases
API price $2/$6 · each benchmark row carries its own source badge (see methodology)