SShogo Bench

Grok 4.5 vs Grok 4.6

Grok 4.6 leads Grok 4.5 by 6.8 points on the Shogo Agent Index, but their ranges overlap, so treat it as close. Grok 4.5 wins 1 and Grok 4.6 wins 19 of 20 shared benchmarks. Grok 4.5 is cheaper for agent workloads.

Add more models to this comparison →

Pillars

PillarGrok 4.5Grok 4.6
Agentic coding5461
Long-horizon autonomy5968
Knowledge-work agents4247
Computer use3642
Security & context learning4852

Shared benchmarks

BenchmarkGrok 4.5Grok 4.6
APEX-Agentsindex56.2%65.3%
DeepSWEindex53.8%67.5%
Vending-Bench 2index$3,887$9,047
ARC-AGI87.2%87.5%
ARC-AGI-252.6%67.1%
Chess Puzzles35.0%33.0%
CritPt15.4%19.7%
DTBench96.5%97.3%
FrontierCode42.4%48.0%
FrontierMath Tier 424.4%31.7%
FrontierMath Tiers 1-357.2%66.0%
GPQA Diamond93.4%94.0%
LMCA45.2%48.5%
MirrorCode32.2%41.1%
OTIS Mock AIME97.8%99.2%
ProofBench31.0%51.0%
SciCode55.0%56.5%
SimpleBench70.0%75.9%
SimpleQA Verified48.3%49.3%
WebDev Arena15551618

Specs and price

Grok 4.5Grok 4.6
Agent Index49.4 (#36)56.2 (#22)
Agent $/M$0.81$0.98
Input / output $/M$2.00 / $6.00$2.00 / $6.00
Context500K500K
ReleasedJul 8, 2026Aug 12, 2026
WeightsClosedClosed
API hosts44

Don’t pick one. Route.

Shogo can run Grok 4.5 and Grok 4.6 side by side and send each step of an agent to the one that’s cheaper for it.

Try both in Shogo