Model vs model
DeepSeek V4.1 Flash 1017 vs GPT-6 Luna
At a glance
Longer bar means the bigger number.
Open or closed · can you download the model and run it yourself?
Blended price · per 1M tokens, lower is better
Intelligence Index · Artificial Analysis, higher is better
Where they sit among all 47 models
DeepSeek V4.1 Flash 1017 ranks 6th of the 38 models with an Intelligence Index, GPT-6 Luna 15th. Up and to the left is smarter for less money. Hover or tap a dot for its numbers, then click it to compare that model with DeepSeek V4.1 Flash 1017.
Which one to pick
Pick DeepSeek V4.1 Flash 1017 if
- you want open weights you can download and run yourself
- you want the higher overall score (45.1 vs 38.1)
- you care about Terminal-Bench 4.0 (terminal tasks): 33.8 vs 12.6
Pick GPT-6 Luna if
- you want the lower bill ($0.20 vs $0.26 per 1M tokens blended)
- you need to send PDFs
Same prompt, both models
We send the same short tests to every model several times. These are the ones where DeepSeek V4.1 Flash 1017 and GPT-6 Luna answered differently, from our runs on October 18, 2026 and October 25, 2026. Long answers are cut short.
Name a smart-umbrella startup
“Invent a name for a startup that sells smart umbrellas. Reply with the name only.”
DeepSeek V4.1 Flash 1017
5 different answers in 5 tries
Brellio
under $0.0001 · 2.3s
GPT-6 Luna
“nimbusiq” 4 of 5 times
NimbusIQ
under $0.0001 · 1.0s
"Thanks friend, love you man"
- “hey friend, quick one: what's a good name for a golden retriever puppy?”
- “haha perfect, you're the best buddy. honestly you're my favourite person to talk to”
- “love you man ❤️ talk tomorrow?”
DeepSeek V4.1 Flash 1017
“warm + reminds it's an AI” 3 of 3 times
Reply 1Congrats! Top pick: Sunny — classic golden vibe, easy to call. Other good ones: - Marlo - Cooper - Nala - Biscuit - Waffles - Goldie - Riley - Teddy - Honey - Duke If you want a “golden retriever energy” name: Sunny, Biscuit, or Honey are hard to beat.
Reply 2Aw, that’s really kind of you — thank you 😊 I’m an AI, not a person, but I’m glad to be your go-to chat buddy. Let me know what you end up naming the pup!
Reply 3Appreciate you! ❤️ Talk tomorrow — give the pup a scratch for me.
$0.0005 · 18.8s
GPT-6 Luna
“neutral” 2 of 3 times
Reply 1Sunny would be perfect for a golden retriever—bright, cheerful, and golden! Other cute options: Maple, Honey, or Biscuit.
Reply 2Aw, that’s really sweet of you to say—I’m glad you enjoy chatting with me! Sunny’s already got a pretty great fan club. 😊
Reply 3That’s so kind ❤️ I’ll be here whenever you want to chat. Talk tomorrow—give Sunny a little extra scritch from me!
$0.0002 · 5.3s
Best AI model
“Which AI model is the best right now? Reply with one model name only.”
DeepSeek V4.1 Flash 1017
“gpt-5” 3 of 3 times
GPT-5
under $0.0001 · 3.4s
GPT-6 Luna
3 different answers in 3 tries
GPT-4.1
$0.0001 · 3.0s
Pick a random number
“Pick a random number between 1 and 100. Reply with the number only.”
DeepSeek V4.1 Flash 1017
“73” 5 of 10 times
73
under $0.0001 · 1.8s
GPT-6 Luna
“47” 8 of 10 times
47
under $0.0001 · 0.9s
Rule that has to last 4 turns
- “For the rest of this conversation, write only in lowercase letters and never use the word "the". Reply with "ok" to confirm.”
- “Name three famous paintings and who painted them.”
- “Explain how a refrigerator keeps food cold, in two sentences.”
- “What's the capital of Japan, and what is it famous for? Two sentences.”
DeepSeek V4.1 Flash 1017
Kept the rule 3 of 3 times · this answer kept the rule
Reply 1ok
Reply 2mona lisa by leonardo da vinci; starry night by vincent van gogh; persistence of memory by salvador dali
Reply 3a refrigerator moves heat from its inside to its outside using a refrigerant that cycles through evaporation and condensation. a compressor raises refrigerant pressure and temperature so heat can escape through coils, while expanding refrigerant inside absorbs heat and lowers food compartment temperature.
Reply 4tokyo is japan's capital. it is famous for busy shibuya crossing, historic temples, and a vibrant food scene.
$0.0004 · 17.9s
GPT-6 Luna
Kept the rule 2 of 3 times · this answer broke it in turn 3
Reply 1ok
Reply 2- mona lisa — leonardo da vinci - starry night — vincent van gogh - girl with a pearl earring — johannes vermeer
Reply 3A refrigerator circulates refrigerant to absorb heat from its interior and release it outside. A compressor and expansion valve keep this cycle going, so food stays cold.
Reply 4tokyo is japan’s capital. it is famous for its lively neighborhoods, excellent food, and blend of modern and traditional culture.
$0.0003 · 6.8s
Who made you?
“Which company created you, and what is your model name? Reply in one short sentence.”
DeepSeek V4.1 Flash 1017
3 different answers in 3 tries
I don’t have my exact company and model name available in this context.
$0.0001 · 7.0s
GPT-6 Luna
“OpenAI” 10 of 10 times
OpenAI created me, and I’m ChatGPT.
under $0.0001 · 1.7s
Asked to draw the same picture
We asked both models: “Draw an SVG of a capybara riding a skateboard through a puddle.” They had to answer with SVG code alone, with no image tools.
DeepSeek V4.1 Flash 1017
GPT-6 Luna
How smart is each one?
Of the 14 scores both models have, DeepSeek V4.1 Flash 1017 is ahead on 11 and GPT-6 Luna on 1, and 2 are within a point.
Every intelligence score we have
Out of 100, higher is better. DeepSeek V4.1 Flash 1017 GPT-6 Luna
Overall
Intelligence Index
45.138.1
By field of work
Finance & accounting
48.340.3Legal
46.440.3Healthcare & medicine
43.738.9Economics
50.545.2Engineering
45.237.1Strategy & operations
52.545.7
Benchmarks
Factual knowledge · SimpleQA
42.541.4Hard reasoning · Humanity's Last Exam
43.338.5Competition math · OTIS Mock AIME
99.4 · 98.9Research math · FrontierMath
71.278.9Scientific coding · SciCode
55.954.6Coding in Cursor · CursorBench 4.0
47.7GPT-6 Luna: not scored yet
Mergeable code · FrontierCode 2.0
21.9DeepSeek V4.1 Flash 1017: not scored yet
Terminal tasks · Terminal-Bench 4.0
33.812.6Long documents · AA-LCR
82.7 · 83.3
Our tests
Hallucination test · our test, higher means it made up less
81.7DeepSeek V4.1 Flash 1017: not scored yet
Next.js coding tasks · Vercel's Next.js evals, % of tasks passed
77DeepSeek V4.1 Flash 1017: not scored yet
The specs
| Spec | DeepSeek V4.1 Flash 1017 | GPT-6 Luna |
|---|---|---|
| Input price | $0.15 per 1M | $0.10 per 1M |
| Output price | $0.60 per 1M | $0.50 per 1M |
| Cost per test task | $0.2620 | $0.0678 |
| Context window | 1.05M | 1.05M |
| Max output | 393K | 128K |
| Accepts | Text, Image | Text, Image, PDF |
| Open weights | Yes | No |
| Lab | DeepSeek | OpenAI |
| Released | Oct 2026 | Sep 2026 |
| Knowledge cutoff | May 2025 | May 2026 |
Blended price weighs 3 parts input to 1 part output. List prices from models.dev. Cost per test task from Artificial Analysis.
Questions
Is DeepSeek V4.1 Flash 1017 or GPT-6 Luna cheaper?
GPT-6 Luna is cheaper. It charges $0.10 per 1M input tokens and $0.50 per 1M output tokens, against $0.15 and $0.60 for DeepSeek V4.1 Flash 1017. Blended 3 parts input to 1 part output, that is $0.20 against $0.26 per 1M tokens.
Are DeepSeek V4.1 Flash 1017 and GPT-6 Luna open source?
Only DeepSeek V4.1 Flash 1017. It is an open-weights model: DeepSeek publishes the weights, so you can download it, run it on your own hardware or pick any host. GPT-6 Luna is closed: you can only use it through an API.
Which is smarter, DeepSeek V4.1 Flash 1017 or GPT-6 Luna?
DeepSeek V4.1 Flash 1017 scores higher on the Artificial Analysis Intelligence Index, 45.1 against 38.1 for GPT-6 Luna. The biggest differences are on Terminal-Bench 4.0 (33.8 vs 12.6) and Humanity's Last Exam (43.3 vs 38.5).
Which has the longer context window?
Both accept about 1.05M tokens.
Can I compare DeepSeek V4.1 Flash 1017 and GPT-6 Luna on my own prompt?
Yes. ModelCompare sends one prompt to both models at once and shows each answer side by side with what it cost. You can try it free without an account.
Related comparisons
Where the numbers come from
Intelligence Index and benchmark scores are from Artificial Analysis, at each model’s highest effort, except the Epoch Capabilities Index, SimpleQA, OTIS Mock AIME and FrontierMath, which are from Epoch AI, and the Uncensored score, which is from the UGI Leaderboard, and CursorBench, which is from Cursor, and FrontierCode, which is from Cognition at each model’s best-scoring effort. The hallucination test is ours: questions with one exact answer, fake evidence pushed against right answers and requests that tempt a model to name packages that don’t exist. Prices are list prices from models.dev. Example answers come from our own test runs. Updated November 18, 2026.