Updated September 27, 2026. Rank says GPT-5.4 is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.
Both of these models will change. Get the price, version and retirement notices for the pair, each with its source. Follow model changes
GPT-5.4 has the higher public score estimate, 68.51 versus 63.48, but the 90% score intervals overlap. Treat that as a lead, not a settled winner. 11 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.
Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.
Tool use, computer use, and multi-step task completion
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 leads on the public agentic lane, 55 to 49.4, with Supported evidence for both models, although the 90% intervals overlap.
Prompts that approach the documented context limit
GPT-5.4
GPT-5.4 has the larger documented context window.
1K fresh input + 500 output tokens
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
200K cached + 20K fresh input + 10K output tokens
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
50K fresh input + 3K output tokens
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Code generation, repair, and software-engineering tasks
No clear pick
The like-for-like coding result is a practical tie on the public lane (within 0.5 points).
Choose a job from the LLM Selector's task catalog to see the category row it rests on, under the same basis rules as the table below. A directional row stays directional; choosing a job never creates a winner.
The same task catalog as the LLM Selector. Each job names the evidence surface it rests on; nothing here adds a new score.
Like-for-like · BenchAlign v5.7
The like-for-like coding row is a practical tie.
Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
1 category rests on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.
Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.
Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.
ARC-AGI-2Reasoning
Normalized gap 12.7HLEKnowledge
Normalized gap 9.4Terminal-Bench 2.0Agentic
Normalized gap 7.2GPQAKnowledge
Normalized gap 2.7SWE-bench ProCoding
Normalized gap 2.3Each row shows the public-lane category score for both models: the BenchAlign v5.7 lane for agentic, coding, and knowledge, and the weighted public rows elsewhere, with the same rank each model page reports. A row is like-for-like only when both scores rest on Supported evidence or the same weighted set. Directional and not-comparable rows remain visible, but they never receive a winner.
| Category | DeepSeek V4 Pro 0813 | GPT-5.4 | Basis | Reading |
|---|---|---|---|---|
| Agentic | 55.0Supported · #29/105 | 49.4Supported · #36/105 | Like-for-likeBenchAlign v5.7 lane · 11 vs 14 public rows | DeepSeek V4 Pro 0813 leads · intervals overlap |
| Coding | 50.3Supported · #39/135 | 50.7Supported · #38/135 | Like-for-likeBenchAlign v5.7 lane · 15 vs 4 public rows | Practical tie |
| Knowledge | 63.4Estimated · #29/158 | 66.8Supported · #16/158 | Directional onlyBenchAlign v5.7 lane · 8 vs 7 public rows | Directional only |
| Reasoning | 56.9Unranked · 4 rankable rows | 60.5#14/19 | Not comparableProvisional lane · 1 vs 2 weighted rows | Not comparable |
| Multimodal | Not ranked | 69.3#21/50 | Not comparableProvisional lane · 0 vs 3 weighted rows | Not comparable |
| Multilingual | Not ranked | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
| Instruction following | Not ranked | 89.2#19/124 | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
| Math | 80.2Unranked · 4 rankable rows | 64.4Unranked · 2 rankable rows | Not comparableProvisional lane · 1 vs 2 weighted rows | Not comparable |
Ranks count the models scored in each category’s lane, so the agentic, coding, and knowledge denominators (BenchAlign v5.7) differ from the provisional-lane categories. Unranked scores sit on the provisional lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.
Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.
1K fresh input + 500 output tokens
DeepSeek V4 Pro 0813 has the lower modeled cost
Costs use the listed standard API rates.
50K fresh input + 3K output tokens
DeepSeek V4 Pro 0813 has the lower modeled cost
Costs use the listed standard API rates.
200K cached + 20K fresh input + 10K output tokens
DeepSeek V4 Pro 0813 has the lower modeled cost
Costs use the listed standard API rates.
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
Maximum documented context; output-token limits may be lower.
DeepSeek V4 Pro 0813
GPT-5.4
1.05M
OpenAI pricingDeepSeek V4 Pro 0813
deepseek-v4-pro
DeepSeek models and pricingGPT-5.4
gpt-5.4
OpenAI pricingA missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
DeepSeek V4 Pro 0813
$0.044 per 1M cached input tokens
DeepSeek: Models & PricingGPT-5.4
$0.25 per 1M cached input tokens
OpenAI pricingDeepSeek V4 Pro 0813
GPT-5.4
Not sourced
DeepSeek V4 Pro 0813
GPT-5.4
Not sourced
DeepSeek V4 Pro 0813
Generally Available · DeepSeek API
DeepSeek models and pricing, V4 Pro continuation footnoteGPT-5.4
Not sourced
DeepSeek V4 Pro 0813
Reasoning
GPT-5.4
Reasoning
DeepSeek V4 Pro 0813
Open Weight
GPT-5.4
Proprietary
DeepSeek V4 Pro 0813
Open Weight
GPT-5.4
Proprietary
DeepSeek V4 Pro 0813
2026-08-13
GPT-5.4
2026-03-05
GPT-5.4 has the higher public score estimate, 68.51 versus 63.48, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.
The like-for-like coding row is a practical tie on the public lane, 50.3 against 50.7, inside the 0.5-point band BenchLM treats as level.
DeepSeek V4 Pro 0813 leads the public agentic tasks lane, 55 to 49.4, with Supported evidence for both models, although the 90% intervals overlap.
For the stated presets, chat costs $0.0033 on DeepSeek V4 Pro 0813 and $0.01 on GPT-5.4; repository review costs $0.07788 and $0.17; the cache-heavy agent loop costs $0.0748 and $0.25. Costs use the listed standard API rates.
GPT-5.4 has the larger documented context window: 1.05M, compared with 1M.
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Terminal-Bench 2.0
GPT-5.4 leads this result
Terminal-Bench 2.1
Not directly comparable
BrowseComp
DeepSeek V4 Pro 0813 leads this result
HLE w/ tools
Not directly comparable
MCP Atlas
DeepSeek V4 Pro 0813 leads this result
Toolathlon
GPT-5.4 leads this result
CyberGym
DeepSeek V4 Pro 0813 leads this result
Toolathlon-Verified
Not directly comparable
Agents' Last Exam
Not directly comparable
AutomationBench
Not directly comparable
Terminal-Bench 2.1 (Vals)
Not directly comparable
OSWorld-Verified
Not directly comparable
τ²-bench results
Not directly comparable
Claw-Eval
Not directly comparable
DeepSearchQA
Not directly comparable
Gert Labs
Not directly comparable
ResearchClawBench
Not directly comparable
JobBench
Not directly comparable
ExploitGym
Not directly comparable
ApprenticeBench
Not directly comparable
LiveCodeBench Pass@1-COT
Not directly comparable
Codeforces
Not directly comparable
SWE-bench Verified
Not directly comparable
SWE-bench Pro
GPT-5.4 leads this result
SWE Multilingual
Not directly comparable
Terminal-Bench 2.0
Not directly comparable
Vibe Code Bench
Shared sourceGPT-5.4 leads this result
Terminal-Bench 2.1
Not directly comparable
NL2Repo
Not directly comparable
DeepSWE
Not directly comparable
DSBench-FullStack
Not directly comparable
DSBench-Hard
Not directly comparable
OpenHarmony Bench
Not directly comparable
LiveCodeBench (Vals)
Not directly comparable
SWE-bench (Vals)
Not directly comparable
LiveCodeBench Pro
Not directly comparable
React Native Evals
Not directly comparable
MRCR 1M
Not directly comparable
CorpusQA 1M
Not directly comparable
ARC-AGI-1
Not directly comparable
ARC-AGI-2
GPT-5.4 leads this result
ARC-AGI-3
Not directly comparable
MMMU-Pro
Not directly comparable
OfficeQA Pro
Not directly comparable
MMMU-Pro w/ Python
Not directly comparable
CharXiv
Not directly comparable
ERQA
Not directly comparable
SimpleVQA
Not directly comparable
ScreenSpot Pro
Not directly comparable
ZeroBench
Not directly comparable
MedXpertQA (MM)
Not directly comparable
MMLU-Pro
Not directly comparable
SimpleQA
Not directly comparable
Chinese-SimpleQA
Not directly comparable
GPQA
GPT-5.4 leads this result
GPQA-D
GPT-5.4 leads this result
HLE
GPT-5.4 leads this result
GPQA Diamond (Vals)
Not directly comparable
MMLU-Pro (Vals)
Not directly comparable
HLE w/o tools
Not directly comparable
HealthBench Hard
Not directly comparable
MedXpertQA (Text)
Not directly comparable
HealthBench Professional
Not directly comparable
HMMT Feb 2026
Not directly comparable
IMOAnswerBench
Not directly comparable
Apex
Not directly comparable
Apex Shortlist
Not directly comparable
FrontierMath v2 (Tiers 1-3)
Not directly comparable
FrontierMath v2 (Tier 4)
Not directly comparable
One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.
Read a sample issueJoin 2,000+ readers.
Last updated September 27, 2026
