Model comparison
DeepSeek-V3.2-Exp vs Gemini 3.1 Flash Lite
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.9 to 40.4 on the Noometry Index.
Last verified . 31 shared benchmarks.
DeepSeek
GoogleSummary
- They share 31 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 7 categories and Gemini 3.1 Flash Lite in 2 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V3.2-Exp leads 51.6 to 41.7.
- The biggest single-benchmark swing is NYT Connections (extended): 36.7% for DeepSeek-V3.2-Exp and 8.2% for Gemini 3.1 Flash Lite.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $0.25 / $1.50 for Gemini 3.1 Flash Lite.
- Gemini 3.1 Flash Lite accepts more context: 1.05M tokens versus 164K.
- DeepSeek-V3.2-Exp has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3.2-Exp | Gemini 3.1 Flash Lite | |
|---|---|---|
| Provider | DeepSeek | |
| Noometry Index | 44.9 | 40.4 |
| Released | 2025-09-29 | 2026-03-03 |
| Weights | Open | Proprietary |
| Context window | 164K | 1.05M |
| Max output | 66K | 66K |
| Input $ / M tokens | $0.26 | $0.25 |
| Output $ / M tokens | $0.38 | $1.50 |
| Results tracked | 48 | 38 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 46.4 (#61), Gemini 3.1 Flash Lite: 37.5 (#179)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3.1 Flash Lite |
|---|---|---|
| LMArena WebDev | 1362 | 1256 |
| SciCode | 38.9% | 41.9% |
| WeirdML | 39.5% | 52.2% |
| LMArena Coding | 1454 | 1400 |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| SWE-bench Multilingual | 59% | — |
| ALE-Bench | — | 797.73 |
Agentic & Tool Use DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 37.9 (#31), Gemini 3.1 Flash Lite: 29.9 (#72)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3.1 Flash Lite |
|---|---|---|
| Terminal-Bench | 39.6% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| DeepResearch Bench | — | 37.3% |
| Vending-Bench 2 | 1,034 | — |
Reasoning Too close to call
DeepSeek-V3.2-Exp: 22.0 (#185), Gemini 3.1 Flash Lite: 22.3 (#181)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3.1 Flash Lite |
|---|---|---|
| Kagi LLM Benchmark | 52.2% | 67.2% |
| NYT Connections (extended) | 36.7% | 8.2% |
| CritPt | 2.9% | 1.1% |
| Chess Puzzles | 14% | 25% |
| Thematic Generalization | 65% | 63.3% |
| LMArena Hard Prompts | 1433 | 1407 |
| DTBench | 87.7% | 76.8% |
| LMCA | 29.1% | 35% |
| Epoch Capabilities Index | 146.28 | 144.45 |
| ARC-AGI-2 | 4% | — |
| ARC-AGI-1 | 57% | — |
| EnigmaEval | — | 3% |
| ForecastBench | — | 54.4 |
Math DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 40.8 (#75), Gemini 3.1 Flash Lite: 39.5 (#82)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3.1 Flash Lite |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 80% |
| LMArena Math | 1435 | 1427 |
| FrontierMath (Tiers 1-3) | — | 27.7% |
| MathArena Final-Answer Competitions | 57.7% | — |
| ProofBench | 8% | — |
| FrontierMath (Feb 2025 set) | 22.1% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 51.6 (#61), Gemini 3.1 Flash Lite: 41.7 (#95)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3.1 Flash Lite |
|---|---|---|
| GPQA Diamond | 83.4% | 81.8% |
| Vectara Hallucination Rate | 5.3% | 8.2% |
| LMArena Expert | 1436 | 1397 |
| Humanity's Last Exam | — | 8.6% |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, Gemini 3.1 Flash Lite: 38.7 (#57)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3.1 Flash Lite |
|---|---|---|
| LMArena Vision | — | 1240 |
Multilingual Too close to call
DeepSeek-V3.2-Exp: 52.4 (#85), Gemini 3.1 Flash Lite: 52.5 (#82)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3.1 Flash Lite |
|---|---|---|
| LMArena Non-English | 1409 | 1410 |
| LMArena Chinese | 1461 | 1461 |
| LMArena French | 1435 | 1423 |
| LMArena German | 1440 | 1429 |
| LMArena Japanese | 1377 | 1411 |
| LMArena Korean | 1372 | 1393 |
| LMArena Russian | 1423 | 1420 |
| LMArena Spanish | 1441 | 1420 |
Instruction Following DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 74.7 (#89), Gemini 3.1 Flash Lite: 72.9 (#126)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3.1 Flash Lite |
|---|---|---|
| LMArena Instruction Following | 1413 | 1376 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#15), Gemini 3.1 Flash Lite: 42.5 (#117)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3.1 Flash Lite |
|---|---|---|
| LMArena Longer Query | 1428 | 1394 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 62.5 (#73), Gemini 3.1 Flash Lite: 61.0 (#90)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3.1 Flash Lite |
|---|---|---|
| LMArena Text | 1425 | 1416 |
| LMArena Creative Writing | 1403 | 1401 |
| LMArena Multi-Turn | 1427 | 1417 |
| EQ-Bench Creative Writing | 1515 | — |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than Gemini 3.1 Flash Lite?
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.9 to 40.4 on the Noometry Index.
Which is cheaper, DeepSeek-V3.2-Exp or Gemini 3.1 Flash Lite?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; Gemini 3.1 Flash Lite lists at $0.25 and $1.50.
Is DeepSeek-V3.2-Exp or Gemini 3.1 Flash Lite better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.4 versus 37.5 in the Noometry coding category.
Which has the bigger context window?
Gemini 3.1 Flash Lite does, with 1.05M tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and Gemini 3.1 Flash Lite share?
31 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 48 scored results on Noometry and Gemini 3.1 Flash Lite has 38.