Back to Developer ZoneOfficial Pricing Official Pricing Official Pricing Official Pricing Official Pricing Official Pricing Official Pricing Official Pricing Official Pricing Official Pricing Official Pricing Official Pricing
OpenAI Models
Explore all 12 models from OpenAI with detailed pricing, pros & cons, and developer recommendations.
12
Models
$0.100
Lowest Input
1.0M
Max Context
4
Quality Tiers
Quick Recommendations
Best Value: GPT-4.1 Nano ($0.100/1M)
Best Quality: GPT-5.5
Best for Reasoning: GPT-5.5 Pro
GPT-5.5
FlagshipAgentic coding, complex workloads
When to use: Best for complex coding agents, computer use, and professional workloads where accuracy matters most.
Upgrade Highlights
- ◆SWE-bench Verified: 82.7% (+6.5pp vs GPT-5.4's 76.2%)
- ◆Terminal-Bench: 82.7% — new SOTA for agentic coding
- ◆1M context window retained, 33% fewer tokens vs GPT-5.4 at same latency
- ◆GeneBench scientific reasoning: 28.5% (+3.2pp over GPT-5.4)
- ◆Price: $5/$30 — 2x GPT-5.4, but cached input 90% cheaper at $0.50/M
Input Price
$5.00
per 1M tokens
Output Price
$30.00
per 1M tokens
Cached Input
$0.500
per 1M tokens
Batch Input
$2.50
per 1M tokens
Context Window: 1M
Max Output: 128,000 tokens
Knowledge Cutoff: 2025-06
VisionFunction CallingFine-tuningJSON ModeFree Tier
Pros
- State-of-the-art agentic coding (82.7% Terminal-Bench)
- 1M token context window
- Matches GPT-5.4 latency with fewer tokens
- Strong scientific research capability
Cons
- 2x more expensive than GPT-5.4
- No fine-tuning support yet
- Output costs $30/M tokens
Performance
Output Speed~85 tok/s
Rate Limit5,000 RPM
Multimodal
Image InputImage OutputAudio InputAudio Output
Benchmarks
MMLU
90.0%
SWE-bench Verified
82.7%
Terminal-Bench
82.7%
GeneBench
28.5%
GPT-5.5 Pro
ReasoningPremium accuracy, parallel compute
When to use: Use when you need maximum accuracy for legal, medical, or scientific analysis and cost is secondary.
Upgrade Highlights
- ◆Parallel test-time compute: runs multiple reasoning paths, picks best
- ◆GeneBench (scientific reasoning): 33.2% — highest ever recorded
- ◆Dedicated GPU allocation for Pro/Business/Enterprise subscribers
- ◆Same 1M context + 128K output as GPT-5.5 base, but deeper reasoning
- ◆6x price premium ($30/$180) — justified only for mission-critical analysis
Input Price
$30.00
per 1M tokens
Output Price
$180.00
per 1M tokens
Cached Input
—
per 1M tokens
Batch Input
—
per 1M tokens
Context Window: 1M
Max Output: 128,000 tokens
Knowledge Cutoff: 2025-06
VisionFunction CallingFine-tuningJSON Mode
Pros
- Highest accuracy via parallel test-time compute
- 33.2% on GeneBench (scientific reasoning)
- Dedicated GPU allocation for Pro subscribers
Cons
- 6x more expensive than GPT-5.5 base
- No batch API support
- Only for Pro/Business/Enterprise tiers
Performance
Output Speed~40 tok/s
Rate Limit3,000 RPM
Multimodal
Image InputImage OutputAudio InputAudio Output
Benchmarks
MMLU
90.5%
GeneBench
33.2%
SWE-bench Verified
79.0%
GPT-5.4
FlagshipCoding, computer use, knowledge work
When to use: Excellent all-rounder for coding assistants, desktop automation, and enterprise knowledge work.
Upgrade Highlights
- ◆OSWorld: 75% — first AI to exceed human-level desktop automation
- ◆33% fewer factual errors vs GPT-5.2 (hallucination reduction)
- ◆Tool Search: dynamically discovers tools, cuts agent token usage by ~40%
- ◆Context: 128K → 1M tokens (8x increase from GPT-4o)
- ◆Max output: 16K → 128K tokens (8x increase), price $2.50/$15
Input Price
$2.50
per 1M tokens
Output Price
$15.00
per 1M tokens
Cached Input
$0.250
per 1M tokens
Batch Input
$1.25
per 1M tokens
Context Window: 1M
Max Output: 128,000 tokens
Knowledge Cutoff: 2025-03
VisionFunction CallingFine-tuningJSON ModeFree Tier
Pros
- First AI to exceed human desktop performance (75% OSWorld)
- Unified coding + computer use + knowledge work
- 33% fewer factual errors than GPT-5.2
- Tool Search reduces token usage for agents
Cons
- More expensive than GPT-4.1
- No fine-tuning yet
- Input pricing doubles above 272K context
Performance
Output Speed~90 tok/s
Rate Limit5,000 RPM
Multimodal
Image InputImage OutputAudio InputAudio Output
Benchmarks
MMLU
89.5%
SWE-bench Verified
76.2%
OSWorld
75.0%
IFEval
87.0%
GPT-5.4 Mini
Mid-tierCost-effective coding & dev tasks
When to use: Best value for high-volume coding tasks, chatbots, and lighter development workloads.
Upgrade Highlights
- ◆SWE-bench Pro: 54.38% — only 3.3pp below GPT-5.4 Standard's 57.7%
- ◆6x cheaper than GPT-5.4 Standard ($0.75 vs $5/M input)
- ◆400K context window — sufficient for most coding tasks
- ◆Same 128K max output as flagship, 90% cached input savings
- ◆Released 12 days after Standard — fastest mini variant ever
Input Price
$0.750
per 1M tokens
Output Price
$4.50
per 1M tokens
Cached Input
$0.075
per 1M tokens
Batch Input
$0.375
per 1M tokens
Context Window: 400K
Max Output: 128,000 tokens
Knowledge Cutoff: 2025-03
VisionFunction CallingFine-tuningJSON ModeFree Tier
Pros
- 54.38% SWE-bench Pro (close to Standard's 57.7%)
- 6x cheaper than GPT-5.4 Standard
- 400K context window
- Free tier access available
Cons
- Lower reasoning quality than Standard
- Context window smaller than full 1M
- Released 12 days after Standard
Performance
Output Speed~120 tok/s
Rate Limit10,000 RPM
Multimodal
Image InputImage OutputAudio InputAudio Output
Benchmarks
SWE-bench Pro
54.38%
MMLU
86.0%
HumanEval
88.2%
GPT-5.4 Nano
LiteEdge, embedded, mobile
When to use: Ideal for on-device assistants, mobile apps, and scenarios where network latency is a concern.
Upgrade Highlights
- ◆Designed for edge/mobile/IoT — optimized for on-device inference
- ◆128K max output despite 272K context — highest output/context ratio
- ◆Vision + function calling at $0.20/M input — previously flagship-only
- ◆Latency optimized for real-time mobile interactions
- ◆Context 272K vs 1M for Standard — trade-off for speed and cost
Input Price
$0.200
per 1M tokens
Output Price
$1.25
per 1M tokens
Cached Input
—
per 1M tokens
Batch Input
—
per 1M tokens
Context Window: 272K
Max Output: 128,000 tokens
Knowledge Cutoff: 2025-03
VisionFunction CallingFine-tuningJSON ModeFree Tier
Pros
- Ultra-low cost for edge deployments
- Designed for mobile and IoT devices
- 128K max output despite small size
Cons
- Limited benchmarks available
- Smaller context window
- Quality gap for complex tasks
Performance
Output Speed~150 tok/s
Rate Limit15,000 RPM
Multimodal
Image InputImage OutputAudio InputAudio Output
Benchmarks
MMLU
78.5%
HumanEval
76.0%
GPT-4o
FlagshipGeneral purpose, multimodal
When to use: Best default choice for most apps requiring high-quality multimodal output.
Upgrade Highlights
- ◆First natively multimodal model: text + image + audio in one model
- ◆2x faster than GPT-4 Turbo at 50% lower cost ($2.50 vs $5/M input)
- ◆Native audio understanding — no separate Whisper pipeline needed
- ◆Fine-tuning support for domain adaptation (flagship exclusive)
- ◆128K context — superseded by GPT-4.1's 1M for long-document tasks
Input Price
$2.50
per 1M tokens
Output Price
$10.00
per 1M tokens
Cached Input
$1.25
per 1M tokens
Batch Input
$1.25
per 1M tokens
Context Window: 128K
Max Output: 16,384 tokens
Knowledge Cutoff: 2023-10
VisionFunction CallingFine-tuningJSON ModeFree Tier
Pros
- Multimodal (text + image + audio)
- Excellent general-purpose quality
- Strong function calling & JSON mode
Cons
- Higher cost than mini/nano variants
- 128K context vs 1M+ for newer 4.1 models
Performance
Output Speed~75 tok/s
Rate Limit5,000 RPM
Multimodal
Image InputImage OutputAudio InputAudio Output
Benchmarks
MMLU
88.7%
HumanEval
90.2%
MATH
76.6%
IFEval
80.4%
Agents Using This Model
3GPT-4o Mini
LiteFast, cost-efficient tasks
When to use: Great for high-volume, cost-sensitive workloads like chatbots and summarization.
Upgrade Highlights
- ◆Replaced GPT-3.5 Turbo: 46% cheaper, significantly better quality
- ◆Same 128K context as GPT-4o at 1/17th the price
- ◆Scored 82% on MMLU — approaching GPT-4 level on knowledge tasks
- ◆Vision + function calling at $0.15/M — previously flagship-only features
- ◆Fine-tuning available — best cost/quality ratio for tuned models
Input Price
$0.150
per 1M tokens
Output Price
$0.600
per 1M tokens
Cached Input
$0.075
per 1M tokens
Batch Input
$0.075
per 1M tokens
Context Window: 128K
Max Output: 16,384 tokens
Knowledge Cutoff: 2023-10
VisionFunction CallingFine-tuningJSON ModeFree Tier
Pros
- Extremely affordable
- Same 128K context as GPT-4o
- Supports vision + function calling
Cons
- Lower reasoning quality than flagship
- Not ideal for complex multi-step tasks
Performance
Output Speed~130 tok/s
Rate Limit10,000 RPM
Multimodal
Image InputImage OutputAudio InputAudio Output
Benchmarks
MMLU
82.0%
HumanEval
87.2%
MATH
67.5%
GPT-4.1
FlagshipCoding, instruction following
When to use: Top pick for coding assistants and long-document analysis.
Upgrade Highlights
- ◆Context: 128K → 1M tokens (8x jump from GPT-4o)
- ◆SWE-bench Verified: 54.6% — best coding score at launch
- ◆Max output: 32K tokens (2x GPT-4o's 16K)
- ◆Instruction following: 40% better on IFEval vs GPT-4o
- ◆Cheaper than GPT-4o ($2 vs $2.50/M input) with more capability
Input Price
$2.00
per 1M tokens
Output Price
$8.00
per 1M tokens
Cached Input
$0.500
per 1M tokens
Batch Input
$1.00
per 1M tokens
Context Window: 1.0M
Max Output: 32,768 tokens
Knowledge Cutoff: 2024-04
VisionFunction CallingFine-tuningJSON ModeFree Tier
Pros
- 1M token context window
- Best-in-class coding ability
- Cheaper than GPT-4o with better context
Cons
- Slower than mini/nano for simple tasks
- Newer model, less battle-tested
Performance
Output Speed~70 tok/s
Rate Limit5,000 RPM
Multimodal
Image InputImage OutputAudio InputAudio Output
Benchmarks
MMLU
88.3%
SWE-bench Verified
54.6%
IFEval
84.0%
HumanEval
92.0%
Agents Using This Model
81GitHub Copilot AgentCursor AgentBolt.newSWE-AgentPerplexity Deep ResearchCrewAILangGraphAutoGenSuperAGIOpenAI Agents SDKLindy.aiRelevance AIOpenAI OperatorBrowser UseLaVagueIntercom FinSierra AIAda AIDrift (Salesloft)Replit AgentCodeium (Windsurf)Sourcegraph Codyv0 by VercelPhindAiderClineStackBlitz AIQodo (CodiumAI)ElicitYou.com AIConsensus NLPIris.aiSTORM by StanfordMicrosoft Copilot StudioAgentGPTAutoGPTBabyAGIMetaGPTChatDevAgency SwarmFlowisen8n AIHaystackMarvin AIDSPyGuidance AINotion AIHubSpot AI AgentSalesforce EinsteinRelay.appActivepiecesWindmillSkyvernAgentforce BrowserPlaywright MCP AgentZendesk AITidio AIFreshdesk Freddy AIForethoughtUltimate (Zendesk)VoiceflowLandbot AIQuickchat AIKore.aiParloaJulius AIRows AIAkkioHex AIPolymer AITableau AI (Agentforce)Levity AIJasper AICopy.aiWritesonicFraseSurfer SEOAnywordPeppertype.aiSimplified AIInVideo AI
GPT-4.1 Mini
Mid-tierBalanced performance & cost
When to use: Sweet spot for production apps needing long context without flagship costs.
Upgrade Highlights
- ◆1M context at $0.40/M input — 5x cheaper than GPT-4.1 flagship
- ◆Same 1M context + 32K output as flagship for production use
- ◆Full multimodal support (vision, function calling, fine-tuning)
- ◆SWE-bench: 42.8% — strong coding at mid-tier pricing
- ◆Replaced GPT-4o Mini for long-context production workloads
Input Price
$0.400
per 1M tokens
Output Price
$1.60
per 1M tokens
Cached Input
$0.100
per 1M tokens
Batch Input
$0.200
per 1M tokens
Context Window: 1.0M
Max Output: 32,768 tokens
Knowledge Cutoff: 2024-04
VisionFunction CallingFine-tuningJSON ModeFree Tier
Pros
- 1M context at mid-tier price
- Strong coding for the price
- Full multimodal support
Cons
- Quality gap vs flagship for complex reasoning
- Still more expensive than nano
Performance
Output Speed~110 tok/s
Rate Limit10,000 RPM
Multimodal
Image InputImage OutputAudio InputAudio Output
Benchmarks
MMLU
84.5%
SWE-bench Verified
42.8%
HumanEval
86.5%
GPT-4.1 Nano
LiteUltra-low latency, edge devices
When to use: Ideal for real-time classification, extraction, and high-throughput pipelines.
Upgrade Highlights
- ◆Cheapest model with 1M context: $0.10/M input (GPT-3.5 Turbo was $0.50)
- ◆Full feature set at lite tier: vision + function calling + fine-tuning
- ◆Context: 128K → 1M (8x increase over GPT-4o Mini's 128K)
- ◆Fastest response times in OpenAI lineup — sub-200ms typical latency
- ◆Cached input at $0.025/M — 97.5% savings for repeated prefixes
Input Price
$0.100
per 1M tokens
Output Price
$0.400
per 1M tokens
Cached Input
$0.025
per 1M tokens
Batch Input
$0.050
per 1M tokens
Context Window: 1.0M
Max Output: 32,768 tokens
Knowledge Cutoff: 2024-04
VisionFunction CallingFine-tuningJSON ModeFree Tier
Pros
- Cheapest OpenAI model with 1M context
- Fastest response times
- Full feature set (vision, FC, fine-tuning)
Cons
- Noticeable quality drop for complex tasks
- Not suitable for deep reasoning
Performance
Output Speed~160 tok/s
Rate Limit15,000 RPM
Multimodal
Image InputImage OutputAudio InputAudio Output
Benchmarks
MMLU
76.0%
HumanEval
72.8%
o3
ReasoningComplex reasoning, math, science
When to use: Use for problems requiring step-by-step reasoning, math proofs, and scientific analysis.
Upgrade Highlights
- ◆ARC-AGI: 87.5% — near-human on novel abstract reasoning tasks
- ◆Max output: 100K tokens (4x o1's 25K) for deep chain-of-thought
- ◆Vision + function calling — new capabilities vs o1's text-only
- ◆Math Olympiad: Gold medal level on AIME 2024 (93.3%)
- ◆Price: $2/$8 — 1/3 the cost of o1 Pro at $200/M
Input Price
$2.00
per 1M tokens
Output Price
$8.00
per 1M tokens
Cached Input
$0.500
per 1M tokens
Batch Input
$1.00
per 1M tokens
Context Window: 200K
Max Output: 100,000 tokens
Knowledge Cutoff: 2024-05
VisionFunction CallingFine-tuningJSON ModeFree Tier
Pros
- State-of-the-art reasoning ability
- 100K max output for long chain-of-thought
- Excellent at math and science
Cons
- Slower due to thinking time
- No fine-tuning support
- Higher cost for reasoning tokens
Performance
Output Speed~25 tok/s
Rate Limit3,000 RPM
Multimodal
Image InputImage OutputAudio InputAudio Output
Benchmarks
ARC-AGI
87.5%
AIME 2024
93.3%
MATH
96.7%
GPQA
83.3%
Agents Using This Model
1o4-mini
ReasoningReasoning at lower cost
When to use: Best value for reasoning workloads — great for production where o3 is overkill.
Upgrade Highlights
- ◆55% cheaper than o3 ($1.10 vs $2/M input) for reasoning tasks
- ◆Same 100K max output as o3 — no compromise on reasoning depth
- ◆Faster inference than o3 with optimized thinking budget
- ◆Vision + function calling at reasoning tier — previously o3-only
- ◆Cached input: $0.275/M — 75% savings for repeated prompts
Input Price
$1.10
per 1M tokens
Output Price
$4.40
per 1M tokens
Cached Input
$0.275
per 1M tokens
Batch Input
$0.550
per 1M tokens
Context Window: 200K
Max Output: 100,000 tokens
Knowledge Cutoff: 2024-05
VisionFunction CallingFine-tuningJSON ModeFree Tier
Pros
- 55% cheaper than o3 for reasoning
- Same 100K max output
- Faster than o3
Cons
- Less capable than o3 on hardest problems
- No fine-tuning
- Reasoning tokens add hidden cost
Performance
Output Speed~35 tok/s
Rate Limit5,000 RPM
Multimodal
Image InputImage OutputAudio InputAudio Output
Benchmarks
AIME 2024
90.0%
MATH
92.5%
GPQA
78.1%
Agents Using This Model
1Side-by-Side Comparison
| Model | Tier | Input | Output | Context |
|---|---|---|---|---|
| GPT-5.5 | Flagship | $5.00 | $30.00 | 1M |
| GPT-5.5 Pro | Reasoning | $30.00 | $180.00 | 1M |
| GPT-5.4 | Flagship | $2.50 | $15.00 | 1M |
| GPT-5.4 Mini | Mid-tier | $0.750 | $4.50 | 400K |
| GPT-5.4 Nano | Lite | $0.200 | $1.25 | 272K |
| GPT-4o | Flagship | $2.50 | $10.00 | 128K |
| GPT-4o Mini | Lite | $0.150 | $0.600 | 128K |
| GPT-4.1 | Flagship | $2.00 | $8.00 | 1.0M |
| GPT-4.1 Mini | Mid-tier | $0.400 | $1.60 | 1.0M |
| GPT-4.1 Nano | Lite | $0.100 | $0.400 | 1.0M |
| o3 | Reasoning | $2.00 | $8.00 | 200K |
| o4-mini | Reasoning | $1.10 | $4.40 | 200K |