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
OGPT-5 Code
Best raw competency at 79.3/100 and the lowest hallucination rate in the set. Pay the premium when accuracy compounds.
If cost matters most
AQwen 3 Coder
90% cheaper than GPT-5, 31.6 points lower on competency. The value pick for high-volume workloads.
OpenAIOGPT-5 Code
Cloud · API
AlibabaAQwen 3 Coder
Open weights
CompetencyWeighted CODE score
79.3/100
flat W21 · rank 2
Best in class
47.7/100
flat W21 · rank 5 · −31.6 vs winner
Cost (input)Per 1M tokens
$14.00
~10× the cheapest in the row
$1.40
90% cheaper than GPT-5
Cheapest
Cost (output)Per 1M tokens
$42.00
~10× the cheapest in the row
$4.20
90% cheaper
Cheapest
SpeedTokens / second
74 tok/s
local inference, batch 1
78 tok/s
local inference, batch 1
Fastest cloud
AccuracyHallucination rate
2.4%
lowest in category
Most accurate
3.6%
+1.2 pp vs winner
ContextMax input tokens
~256 pages
~256 pages context window
Largest
~256 pages
~256 pages context window
DeploymentWhere it runs
Cloud only
API key · quick setup
Cloud or local
Self-host on your GPUs
Most flexible
Data residencyCompliance
Vendor regions
verify region availability
Anywhere
runs inside your perimeter
Most flexible
LicenseWhat you can do
Proprietary
commercial OK · ToS applies
Open weights
commercial OK
Most permissive
Where each one wins.
Sub-competency scores by business task. Bars highlight the row leader in accent.
Code Generation40% of category weight
OGPT-5 Code
75
AQwen 3 Coder
48
Review & Bug Fixing30% weight
OGPT-5 Code
79
AQwen 3 Coder
48
Full-Repo Understanding20% weight
OGPT-5 Code
83
AQwen 3 Coder
48
Test Generation10% weight
OGPT-5 Code
87
AQwen 3 Coder
48
OGPT-5 Code
✓ Best for
Strong at Code Generation for OpenAI.
Strong at Review & Bug Fixing for OpenAI.
⚠ Watch out for
Validate fit for your specific workload before committing.
