vs
0:8
overall capability wins (0 ties across 8 scored). o3 leads overall.
- Reasoning0:3
- Coding0:2
- Arena0:1
- Pricing0:3
- Speed1:0
- Specs0:2
Verdict
o3 leads coding (widest gap: +43.5 on LiveCode); o3 ranks higher on LMArena (+107 Elo); o3 offers ~1.6x larger context (200K vs 128K).
- o3 leads coding (widest gap: +43.5 on LiveCode)
- o3 ranks higher on LMArena (+107 Elo)
- o3 offers ~1.6x larger context (200K vs 128K)
- o3 allows ~24.4x more max output (100K vs 4K)
- o3 is ~4.3x cheaper on a blended token basis than GPT-4 Turbo
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- GPT-4 Turbo
- $17.50
- o3
- $4
o3 is about 4.4x cheaper on this workload.
Price per 1M tokens
Full comparison
Organization
OpenAI
OpenAI
Family
GPT-4 / o-series
GPT-4 / o-series
License
Proprietary
Proprietary
Open weights
No
No
Release
Nov 6, 2023
Apr 16, 2025
Knowledge cutoff
-
-
API / provider
OpenAI
OpenAI
Modalities
text, image → text
text, image → text
Specs
Context window
128K
200K
Max output
4K
100K
Parameters
—
—
Pricing
Input $/1M
$10
$2
Output $/1M
$30
$8
Blended $/1M (3∶1)
$15
$3.50
Speed
tok/s
55
55
TTFT (s)
0.6
1.2
Reasoning
MMLU-Pro
69.4
85.3
GPQA Diamond
—
82.7
Humanity's Last Exam
3.1
20.1
AIME 2025
—
88.9
MATH-500
73.7
99.2
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
—
38.6
AA-LCR v1.1
—
74.7
CritPt
—
1.1
MMMU-Pro
—
70.1
IFBench
—
71.4
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
58.4
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
37.3
80.8
Terminal-Bench 2.1
—
—
Aider Polyglot
—
81.3
Terminal-Bench 3
—
—
BigCodeBench
48.2
—
SciCode
31.9
41.0
CursorBench
—
—
SWE-Rebench
—
38.1
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,324
1,432
| Metric | GPT-4 Turbo | o3 | Delta |
|---|---|---|---|
| Identity | |||
| Organization | OpenAI | OpenAI | — |
| Family | GPT-4 / o-series | GPT-4 / o-series | — |
| License | Proprietary | Proprietary | — |
| Open weights | No | No | — |
| Release | Nov 6, 2023 | Apr 16, 2025 | — |
| Knowledge cutoff | - | - | — |
| API / provider | OpenAI | OpenAI | — |
| Modalities | text, image → text | text, image → text | — |
| Specs | |||
| Context window | 128K | 200K | B +72K |
| Max output | 4K | 100K | B +96K |
| Parameters | — | — | — |
| Pricing | |||
| Input $/1M | $10 | $2 | B +$8 |
| Output $/1M | $30 | $8 | B +$22 |
| Blended $/1M (3∶1) | $15 | $3.50 | B +$11.50 |
| Speed | |||
| tok/s | 55 | 55 | tie |
| TTFT (s) | 0.6 | 1.2 | A +0.60 |
| Reasoning | |||
| MMLU-Pro | 69.4 | 85.3 | B +15.9 pts |
| GPQA Diamond | — | 82.7 | — |
| Humanity's Last Exam | 3.1 | 20.1 | B +17.0 pts |
| AIME 2025 | — | 88.9 | — |
| MATH-500 | 73.7 | 99.2 | B +25.5 pts |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | — | 38.6 | — |
| AA-LCR v1.1 | — | 74.7 | — |
| CritPt | — | 1.1 | — |
| MMMU-Pro | — | 70.1 | — |
| IFBench | — | 71.4 | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | 58.4 | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | 37.3 | 80.8 | B +43.5 pts |
| Terminal-Bench 2.1 | — | — | — |
| Aider Polyglot | — | 81.3 | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | 48.2 | — | — |
| SciCode | 31.9 | 41.0 | B +9.1 pts |
| CursorBench | — | — | — |
| SWE-Rebench | — | 38.1 | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,324 | 1,432 | B +107 Elo |
GPT-4 Turbo
- Context
- 128K / 4K out
- Parameters
- —
- Price
- $10 / $30
- Speed
- 55 tok/s · 0.6s TTFT
- Modalities
- text, image → text
- License
- Proprietary
128K-context GPT-4 refresh with vision — long-running production workhorse before GPT-4o / 4.1.
o3
- Context
- 200K / 100K out
- Parameters
- —
- Price
- $2 / $8
- Speed
- 55 tok/s · 1.2s TTFT
- Modalities
- text, image → text
- License
- Proprietary
OpenAI reasoning model from the o-series; strong on STEM and agentic coding with deliberate chain-of-thought.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.