Long documents
Prompts that approach the documented context limit
DeepSeek V4 Pro 0813
DeepSeek V4 Pro 0813 has the larger documented context window.
Confidence: documented
DeepSeek V3 vs DeepSeek V4 Pro 0813
Updated September 27, 2026. Rank says DeepSeek V4 Pro 0813 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
Share or export
Decision reading
DeepSeek V4 Pro 0813 has the higher public score, 63.48 versus 31.9, and the 90% score intervals do not overlap. 3 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.
- Shared results
- 3
- DeepSeek V3 only
- 3
- DeepSeek V4 Pro 0813 only
- 39
- Like-for-like categories
- 0 / 8
Supported: DeepSeek V3 · Estimated: DeepSeek V4 Pro 0813How the comparison works
Which one for your work
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.
Chat turn cost
1K fresh input + 500 output tokens
DeepSeek V3
DeepSeek V3 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-rates
Show secondary and unsupported calls
Repository review cost
50K fresh input + 3K output tokens
DeepSeek V3
DeepSeek V3 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.
Confidence: listed-ratesCoding work
Code generation, repair, and software-engineering tasks
Not enough matched evidence
DeepSeek V3 is scored on Estimated evidence for coding, so the reading is directional rather than like-for-like.
Confidence: limitedAgentic work
Tool use, computer use, and multi-step task completion
Not enough matched evidence
DeepSeek V3 is scored on Estimated evidence for agentic, so the reading is directional rather than like-for-like.
Confidence: limitedCache-heavy agent loop cost
200K cached + 20K fresh input + 10K output tokens
Not enough matched evidence
The page does not recommend a cost winner because at least one model cannot fit the stated workload in one request. DeepSeek V3 does not fit this workload in one request.
Confidence: listed-rates
Which one for a specific job
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.
18.9DeepSeek V350.3DeepSeek V4 Pro 0813
Directional only · BenchAlign v5.7
DeepSeek V4 Pro 0813 scores higher, but at least one score rests on Estimated evidence or a different benchmark set. Directional only, no winner.
Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.
Same basis rules as the category table below
What is actually comparable
Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.
3 DeepSeek V3 only
3 Shared
39 DeepSeek V4 Pro 0813 only
3 categories rest on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.
Shape of the matched evidence
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.
SWE-bench VerifiedCoding
Normalized gap 38.6DeepSeek V3:42%DeepSeek V4 Pro 0813:80.6%GPQAKnowledge
Normalized gap 31.0DeepSeek V3:59.1%DeepSeek V4 Pro 0813:90.1%MMLU-ProKnowledge
Normalized gap 11.6DeepSeek V3:75.9%DeepSeek V4 Pro 0813:87.5%
Bars run 0–100 on each benchmark’s normalized display scale
Category results, on a stated basis
Each 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.
Agentic
Directional only- DeepSeek V3
- 11.1
- Estimated · #102/105
- DeepSeek V4 Pro 0813
- 55.0
- Supported · #29/105
- Basis
- BenchAlign v5.7 lane · 0 vs 11 public rows
- Reading
- Directional only
Coding
Directional only- DeepSeek V3
- 18.9
- Estimated · #123/135
- DeepSeek V4 Pro 0813
- 50.3
- Supported · #39/135
- Basis
- BenchAlign v5.7 lane · 2 vs 15 public rows
- Reading
- Directional only
Knowledge
Directional only- DeepSeek V3
- 30.0
- Estimated · #130/158
- DeepSeek V4 Pro 0813
- 63.4
- Estimated · #29/158
- Basis
- BenchAlign v5.7 lane · 2 vs 8 public rows
- Reading
- Directional only
Reasoning
Not comparable- DeepSeek V3
- 42.2
- Unranked · 2 rankable rows
- DeepSeek V4 Pro 0813
- 56.9
- Unranked · 4 rankable rows
- Basis
- Provisional lane · 0 vs 1 weighted rows
- Reading
- Not comparable
Multimodal
Not comparable- DeepSeek V3
- Not ranked
- DeepSeek V4 Pro 0813
- Not ranked
- Basis
- Provisional lane · 0 vs 0 weighted rows
- Reading
- Not comparable
Multilingual
Not comparable- DeepSeek V3
- Not ranked
- DeepSeek V4 Pro 0813
- Not ranked
- Basis
- Provisional lane · 0 vs 0 weighted rows
- Reading
- Not comparable
Instruction following
Not comparable- DeepSeek V3
- 38.2
- #103/124
- DeepSeek V4 Pro 0813
- Not ranked
- Basis
- Provisional lane · 0 vs 0 weighted rows
- Reading
- Not comparable
Math
Not comparable- DeepSeek V3
- 26.0
- Unranked · 1 rankable row
- DeepSeek V4 Pro 0813
- 80.2
- Unranked · 4 rankable rows
- Basis
- Provisional lane · 1 vs 1 weighted rows
- Reading
- Not comparable
| Category | DeepSeek V3 | DeepSeek V4 Pro 0813 | Basis | Reading |
|---|---|---|---|---|
| Agentic | 11.1Estimated · #102/105 | 55.0Supported · #29/105 | Directional onlyBenchAlign v5.7 lane · 0 vs 11 public rows | Directional only |
| Coding | 18.9Estimated · #123/135 | 50.3Supported · #39/135 | Directional onlyBenchAlign v5.7 lane · 2 vs 15 public rows | Directional only |
| Knowledge | 30.0Estimated · #130/158 | 63.4Estimated · #29/158 | Directional onlyBenchAlign v5.7 lane · 2 vs 8 public rows | Directional only |
| Reasoning | 42.2Unranked · 2 rankable rows | 56.9Unranked · 4 rankable rows | Not comparableProvisional lane · 0 vs 1 weighted rows | Not comparable |
| Multimodal | Not ranked | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
| Multilingual | Not ranked | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
| Instruction following | 38.2#103/124 | Not ranked | Not comparableProvisional lane · 0 vs 0 weighted rows | Not comparable |
| Math | 26.0Unranked · 1 rankable row | 80.2Unranked · 4 rankable rows | Not comparableProvisional lane · 1 vs 1 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.
Supported evidence per lane · bars run 0–100Methodology
What each workload costs
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.
Chat turn
1K fresh input + 500 output tokens
- DeepSeek V3
- $0.00082
- Fits in one request
- DeepSeek V4 Pro 0813
- $0.0033
- Fits in one request
DeepSeek V3 has the lower modeled cost
Costs use the listed standard API rates.
Repository review
50K fresh input + 3K output tokens
- DeepSeek V3
- $0.0168
- Fits in one request
- DeepSeek V4 Pro 0813
- $0.07788
- Fits in one request
DeepSeek V3 has the lower modeled cost
Costs use the listed standard API rates.
Cache-heavy agent loop
200K cached + 20K fresh input + 10K output tokens
- DeepSeek V3
- $0.0304
- Does not fit in one request
- DeepSeek V4 Pro 0813
- $0.0748
- Fits in one request
DeepSeek V3 does not fit this workload in one request.
Cached input falls back to the list input rate only where a cached rate is unpublished
Specification differences
Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.
SpecificationDeepSeek V3DeepSeek V4 Pro 0813
Context window
Maximum documented context; output-token limits may be lower.
DeepSeek V3
128K
DeepSeek V4 Pro 0813
API model ID
DeepSeek V3
Not sourced
DeepSeek V4 Pro 0813
deepseek-v4-pro
DeepSeek models and pricingCached-input rate
A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.
DeepSeek V3
$0.07 per 1M cached input tokens
DeepSeek V4 Pro 0813
$0.044 per 1M cached input tokens
DeepSeek: Models & PricingDocumented inputs
DeepSeek V3
Not sourced
DeepSeek V4 Pro 0813
Documented outputs
DeepSeek V3
Not sourced
DeepSeek V4 Pro 0813
Provider availability
DeepSeek V3
Not sourced
DeepSeek V4 Pro 0813
Generally Available · DeepSeek API
DeepSeek models and pricing, V4 Pro continuation footnoteReasoning profile
DeepSeek V3
Non-Reasoning
DeepSeek V4 Pro 0813
Reasoning
Weight access
DeepSeek V3
Open Weight
DeepSeek V4 Pro 0813
Open Weight
License
DeepSeek V3
Open Weight
DeepSeek V4 Pro 0813
Open Weight
Release date
DeepSeek V3
2024-12-26
DeepSeek V4 Pro 0813
2026-08-13
If you already use one of these models
- Deployment change
- Both entries list DeepSeek as the provider. Confirm endpoint, model ID, limits, and feature support before switching.
- Quality signal
- DeepSeek V4 Pro 0813 has the higher public score, 63.48 versus 31.9, and the 90% score intervals do not overlap.
- Workload cost
- Repository review: $0.0168 vs $0.07788. Cache-heavy agent loop: $0.0304 vs $0.0748.
- Context tradeoff
- DeepSeek V4 Pro 0813 has the larger documented window (1M).
Run the same representative tasks against both endpoints before changing production traffic.
Questions
Which is better, DeepSeek V3 or DeepSeek V4 Pro 0813?
DeepSeek V4 Pro 0813 has the higher public score, 63.48 versus 31.9, and the 90% score intervals do not overlap. That is the clearest overall quality signal in the current public data.
Which is better for coding, DeepSeek V3 or DeepSeek V4 Pro 0813?
DeepSeek V4 Pro 0813 scores higher for coding on the public lane, 50.3 to 18.9. DeepSeek V3 is scored on Estimated evidence for coding, so the reading is directional rather than like-for-like. BenchLM does not name a winner for coding from a directional reading; read the shared benchmark rows directly and test the models on the same task set.
Which is better for agentic tasks, DeepSeek V3 or DeepSeek V4 Pro 0813?
DeepSeek V4 Pro 0813 scores higher for agentic tasks on the public lane, 55 to 11.1. DeepSeek V3 is scored on Estimated evidence for agentic tasks, so the reading is directional rather than like-for-like. BenchLM does not name a winner for agentic tasks from a directional reading; read the shared benchmark rows directly and test the models on the same task set.
Which costs less, DeepSeek V3 or DeepSeek V4 Pro 0813?
For the stated presets, chat costs $0.00082 on DeepSeek V3 and $0.0033 on DeepSeek V4 Pro 0813; repository review costs $0.0168 and $0.07788; the cache-heavy agent loop costs $0.0304 and $0.0748. DeepSeek V3 does not fit this workload in one request.
Which has the larger context window, DeepSeek V3 or DeepSeek V4 Pro 0813?
DeepSeek V4 Pro 0813 has the larger documented context window: 1M, compared with 128K.
Self-host vs API cost
Estimates at 50,000 req/day · 1000 tokens/req average.
DeepSeek V3
API / mo$1,028
Self-host / mo$18,221
Break-even1.2B/day
DeepSeek V4 Pro 0813
API / mo$0
Self-host / moNot listed
Break-even—
Proprietary model — self-hosting not applicable.
Benchmark evidence
The full public result ledger is available for audit without forcing a wide desktop table onto a phone.
Browse raw public benchmark evidence45 rows
Agentic
Terminal-Bench 2.0
DeepSeek V3—DeepSeek V4 Pro 081367.9%SourceNot directly comparable
Terminal-Bench 2.1
DeepSeek V3—DeepSeek V4 Pro 081387.9%SourceNot directly comparable
BrowseComp
DeepSeek V3—DeepSeek V4 Pro 081383.4%SourceNot directly comparable
HLE w/ tools
DeepSeek V3—DeepSeek V4 Pro 081360.0%SourceNot directly comparable
MCP Atlas
DeepSeek V3—DeepSeek V4 Pro 081373.6%SourceNot directly comparable
Toolathlon
DeepSeek V3—DeepSeek V4 Pro 081351.8%SourceNot directly comparable
CyberGym
DeepSeek V3—DeepSeek V4 Pro 081383.3%SourceNot directly comparable
Toolathlon-Verified
DeepSeek V3—DeepSeek V4 Pro 081374.1%SourceNot directly comparable
Agents' Last Exam
DeepSeek V3—DeepSeek V4 Pro 081325.7%SourceNot directly comparable
AutomationBench
DeepSeek V3—DeepSeek V4 Pro 081331.8%SourceNot directly comparable
Terminal-Bench 2.1 (Vals)
DeepSeek V3—DeepSeek V4 Pro 081354.7%SourceNot directly comparable
Coding
LiveCodeBench
DeepSeek V337.6%SourceDeepSeek V4 Pro 0813—Not directly comparable
SWE-bench Verified
DeepSeek V342%SourceDeepSeek V4 Pro 081380.6%SourceDeepSeek V4 Pro 0813 leads this result
LiveCodeBench Pass@1-COT
DeepSeek V3—DeepSeek V4 Pro 081393.5%SourceNot directly comparable
Codeforces
DeepSeek V3—DeepSeek V4 Pro 08133206.0SourceNot directly comparable
SWE-bench Pro
DeepSeek V3—DeepSeek V4 Pro 081355.4%SourceNot directly comparable
SWE Multilingual
DeepSeek V3—DeepSeek V4 Pro 081376.2%SourceNot directly comparable
Terminal-Bench 2.0
DeepSeek V3—DeepSeek V4 Pro 081367.9%SourceNot directly comparable
Vibe Code Bench
DeepSeek V3—DeepSeek V4 Pro 081349.93%SourceNot directly comparable
Terminal-Bench 2.1
DeepSeek V3—DeepSeek V4 Pro 081387.9%SourceNot directly comparable
NL2Repo
DeepSeek V3—DeepSeek V4 Pro 081361.5%SourceNot directly comparable
DeepSWE
DeepSeek V3—DeepSeek V4 Pro 081362.7%SourceNot directly comparable
DSBench-FullStack
DeepSeek V3—DeepSeek V4 Pro 081371.1%SourceNot directly comparable
DSBench-Hard
DeepSeek V3—DeepSeek V4 Pro 081367.2%SourceNot directly comparable
OpenHarmony Bench
DeepSeek V3—DeepSeek V4 Pro 081359.0%SourceNot directly comparable
LiveCodeBench (Vals)
DeepSeek V3—DeepSeek V4 Pro 081387.5%SourceNot directly comparable
SWE-bench (Vals)
DeepSeek V3—DeepSeek V4 Pro 081396.4%SourceNot directly comparable
Reasoning
MRCR 1M
DeepSeek V3—DeepSeek V4 Pro 081383.5%SourceNot directly comparable
CorpusQA 1M
DeepSeek V3—DeepSeek V4 Pro 081362.0%SourceNot directly comparable
ARC-AGI-1
DeepSeek V3—DeepSeek V4 Pro 081390.00%SourceNot directly comparable
ARC-AGI-2
DeepSeek V3—DeepSeek V4 Pro 081361.3%SourceNot directly comparable
Knowledge
GPQA
DeepSeek V359.1%SourceDeepSeek V4 Pro 081390.1%SourceDeepSeek V4 Pro 0813 leads this result
MMLU-Pro
DeepSeek V375.9%SourceDeepSeek V4 Pro 081387.5%SourceDeepSeek V4 Pro 0813 leads this result
SimpleQA
DeepSeek V3—DeepSeek V4 Pro 081357.9%SourceNot directly comparable
Chinese-SimpleQA
DeepSeek V3—DeepSeek V4 Pro 081384.4%SourceNot directly comparable
GPQA-D
DeepSeek V3—DeepSeek V4 Pro 081390.1%SourceNot directly comparable
HLE
DeepSeek V3—DeepSeek V4 Pro 081342.7%SourceNot directly comparable
GPQA Diamond (Vals)
DeepSeek V3—DeepSeek V4 Pro 081392.4%SourceNot directly comparable
MMLU-Pro (Vals)
DeepSeek V3—DeepSeek V4 Pro 081387.0%SourceNot directly comparable
Instruction following
IFEval
DeepSeek V386.1%SourceDeepSeek V4 Pro 0813—Not directly comparable
Math
FrontierMath v2 (Tiers 1-3)
DeepSeek V31.724%SourceDeepSeek V4 Pro 0813—Not directly comparable
HMMT Feb 2026
DeepSeek V3—DeepSeek V4 Pro 081395.2%SourceNot directly comparable
IMOAnswerBench
DeepSeek V3—DeepSeek V4 Pro 081389.8%SourceNot directly comparable
Apex
DeepSeek V3—DeepSeek V4 Pro 081338.3%SourceNot directly comparable
Apex Shortlist
DeepSeek V3—DeepSeek V4 Pro 081390.2%SourceNot directly comparable
45 public results · 3 shared
Watch DeepSeek V3 vs DeepSeek V4 Pro 0813
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