SGrok 4.6
✓ Best for
⚠ Watch out for
Open the full profile →Compare.
Models head to head — pick two or three. Same scoring, same week. Search to swap, add, or remove a model and the matchup recomputes. Hover or tap a stat to see how it shifted.
If quality matters most
SGrok 4.6
Best raw competency at 93.2/100. Pay the premium when accuracy compounds.
If cost matters most
AClaude Opus 5
-150% cheaper than Grok, 0.1 points lower on competency. The value pick for high-volume workloads.
If you need control
OGPT-5.6 Sol
Self-host inside your perimeter — . Lower quality, but no vendor lock-in and no data leaves your network.
SpaceXAISGrok 4.6
AnthropicAClaude Opus 5
OpenAIOGPT-5.6 Sol
CompetencyWeighted LLM score
93.2/100
flat W21 · rank 1
Best in class93.1/100
flat W21 · rank 2 · −0.1 vs winner
92/100
flat W21 · rank 3 · −1.2 vs winner
Cost (input)Per 1M tokens
$2.00
60% cheaper than Claude
Cheapest$5.00
~3× the cheapest in the row
$4.00
~2× the cheapest in the row
Cost (output)Per 1M tokens
$6.00
76% cheaper
Cheapest$25.00
~4× the cheapest in the row
$20.00
~3× the cheapest in the row
SpeedTokens / second
53 tok/s
TTFT 35.7s
53 tok/s
TTFT 35s
62 tok/s
TTFT 97.6s
Fastest cloudAccuracyHallucination rate
—
no data
—
no data
—
no data
ContextMax input tokens
—
— context window
Largest—
— context window
—
— context window
DeploymentWhere it runs
—
—
—
Most flexibleData residencyCompliance
—
—
—
Most flexibleLicenseWhat you can do
—
Most permissive—
—
Where each one wins.
Sub-competency scores by business task. Bars highlight the row leader in accent.
Customer Service20% of category weight
Content Writing15% weight
Research & Analysis16% weight
Summarization12% weight
Translation7% weight
Reasoning16% weight
Knowledge14% weight
AClaude Opus 5
✓ Best for
⚠ Watch out for
Open the full profile →OGPT-5.6 Sol
✓ Best for
⚠ Watch out for
Open the full profile →Cost vs quality, charted.
The selected models on the value frontier. Up-and-to-the-left wins.
Filled = cloud / APIOutline = open weightsLower-left = best value
Still picking? We can shorten the list.
Ten questions, one answer. Cost estimates, quality bar, and the eval gates we'd add this sprint.