Gemini 3.5 Flash vs Grok 4.6
Full profiles: Gemini 3.5 Flash (Google) and Grok 4.6 (SpaceXAI): every benchmark, price and rank.
Updated · Data from 9 sources
As of 25 Oct 2026, Gemini 3.5 Flash (high) and Grok 4.6 (medium) are neck and neck on quality (63.1 vs 64.2, inside the ±5.1 margin of error) and cost about the same, so speed decides it: Gemini 3.5 Flash (high) streams 2.9× faster. Grok 4.6 leads in every category we measure both in.
Gemini 3.5 Flash benchmarks high Google
Grok 4.6 benchmarks medium SpaceXAIOpen in the comparison toolWhich to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Coding | Grok 4.6 | 62 vs 60 category score; SciCode 55.9% vs 53.9% | Medium |
| Agents & tool use | Grok 4.6 | 74 vs 63 category score; τ³-Banking (AA) 44.3% vs 32.2% | High |
| Reasoning | Grok 4.6 | 64 vs 61 category score; CritPt 17.7% vs 13.1% | Medium |
| Knowledge | Grok 4.6 | 74 vs 65 category score; AA-Omniscience 28 vs 21.2 | High |
| Long documents | Grok 4.6 | 66 vs 62 category score; AA-LCR 81.0% vs 73.3% | Medium |
| Chat quality (human preference) | – | Only Gemini 3.5 Flash has an Arena rating (67 category score) | Not comparable |
| Lowest cost | Either | $3.38 vs $3.00 per 1M blended; agentic coding session $0.126 vs $0.144 | – |
| Fastest first answer | Gemini 3.5 Flash | 14.2 s (Gemini 3.5 Flash (high)) vs 25.5 s (Grok 4.6 (medium)) to the first answer token | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- Gemini 3.5 Flash (high) and Grok 4.6 (medium) are neck and neck on quality (63.1 vs 64.2, within the ±5.1 margin).
- Gemini 3.5 Flash (high) is stronger in human preference, instruction following, maths, multimodal.
- Grok 4.6 (medium) is stronger in agents & tools, coding, composite, knowledge, long context, reasoning.
- They cost about the same ($3.00 per 1M blended).
- Gemini 3.5 Flash (high) streams 2.9× faster (222 vs 78 tokens per second).
At a glance
| Spec | Gemini 3.5 Flash | Grok 4.6 |
|---|---|---|
| Developer | SpaceXAI | |
| Released | 19 May 2026 | 11 Aug 2026 |
| Context window (tokens) | 1,048,576 | 500,000 |
| Compared at | Gemini 3.5 Flash (high) | Grok 4.6 (medium) |
| Open weights | No | No |
| Price in / out / cached input per 1M | $1.50 / $9.00 / $0.150 | $2.00 / $6.00 / $0.500 |
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | Gemini 3.5 Flash (high) | Grok 4.6 (medium) | Leader |
|---|---|---|---|
| Reasoning | 61.3 | 64.4 | Grok 4.6 |
| Coding | 59.8 | 62.3 | Grok 4.6 |
| Agents & tools | 63.4 | 73.5 | Grok 4.6 |
| Maths | 58.3 | – | not comparable |
| Knowledge | 65.2 | 74.1 | Grok 4.6 |
| Instruction following | 74.3 | – | not comparable |
| Human preference | 66.8 | – | not comparable |
| Multimodal | 66.6 | – | not comparable |
| Long context | 61.6 | 65.6 | Grok 4.6 |
| Composite | 65.9 | 77.4 | Grok 4.6 |
Where sources disagree
Across the 17 benchmarks both have, Gemini 3.5 Flash leads 5, Grok 4.6 leads 12. Independent sources split: Gemini 3.5 Flash leads on Arena Text; Grok 4.6 leads on AA Intelligence Index v4.3.2.
| Overall index | Gemini 3.5 Flash (high) | Grok 4.6 (medium) | Leader |
|---|---|---|---|
| BenchLeader Index | 63.1 | 64.2 | neck and neck |
| AA Intelligence Index v4.3.2 | 32.6 | 42.8 | Grok 4.6 |
| Arena Text | 1477 | 1456† | Gemini 3.5 Flash |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.
| Measure | Gemini 3.5 Flash (high) | Grok 4.6 (medium) |
|---|---|---|
| Blended price per 1M tokens | $3.38 | $3.00 |
| Input / output per 1M | $1.50 / $9.00 | $2.00 / $6.00 |
| Cached input per 1M | $0.150 | $0.500 |
| Price per task: chat reply | $0.0033 | $0.0026 |
| Price per task: summarize a 30-page report | $0.023 | $0.028 |
| Price per task: code edit | $0.022 | $0.021 |
| Price per task: agentic coding session | $0.126 | $0.144 |
| Price per task: structured extraction | $0.0048 | $0.0052 |
| Cost to run a full benchmark suite | $3.69 | $1.14 |
| Output speed | 222 tok/s | 78 tok/s |
| Time to first answer | 14.2 s | 25.5 s |
| Time to a complete answer | 16.4 s | 31.9 s |
Per token is not per task: on Artificial Analysis' full suite Gemini 3.5 Flash (high) cost $3.69 and Grok 4.6 (medium) $1.14, because Gemini 3.5 Flash (high) used about 2.9× as many tokens for the same work, which a per-token price hides.
Each model is compared at its best-ranked reasoning-effort setting. Gemini 3.5 Flash (high) scores 63.1 and takes 14 s to its first answer token; Gemini 3.5 Flash (medium) scores 62.8 in 12 s, Gemini 3.5 Flash (minimal) scores 56.4 in 1 s. Grok 4.6 (medium) scores 64.2 and takes 25 s to its first answer token; Grok 4.6 (high) scores 62.9 in 25 s, Grok 4.6 (xhigh) scores 62.5 in 29 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
130 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | Gemini 3.5 Flash (high) | Grok 4.6 (medium) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 63.1 | 64.2 |
| Agents & tools score | BenchLeader | 63.4 | 73.5 |
| Coding score | BenchLeader | 59.8 | 62.3 |
| Composite score | BenchLeader | 65.9 | 77.4 |
| Human preference score | BenchLeader | 66.8 | – |
| Instruction following score | BenchLeader | 74.3 | – |
| Knowledge score | BenchLeader | 65.2 | 74.1 |
| Long context score | BenchLeader | 61.6 | 65.6 |
| Maths score | BenchLeader | 58.3 | – |
| Multimodal score | BenchLeader | 66.6 | – |
| Reasoning score | BenchLeader | 61.3 | 64.4 |
| Blended price $/M | BenchLeader | $3.38 | $3.00 |
| Output speed | BenchLeader | 222 tok/s | 78 tok/s |
| Time to first answer | BenchLeader | 14.2 s | 25.5 s |
| Context window | BenchLeader | 1,048,576 tokens | 500,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 92.8% | 94.0%† |
| SimpleBench | SimpleBench | 76.7%† | 75.9%† |
| Arena Hard Prompts | Arena (LMArena) | 1495 | 1482† |
| LiveBench 2.0 Reasoning | LiveBench | – | 91.2%† |
| LiveBench 1.0 Reasoning | LiveBench | 82.0% | 90.5%† |
| GPQA Diamond (AA) | Artificial Analysis | 92.2% | 93.5% |
| Humanity's Last Exam (AA) | Artificial Analysis | 42.7% | 42.1% |
| GPQA Diamond (Vals) | Vals AI | 92.7% | 94.7%† |
| EnigmaEval | Scale AI SEAL | 25.4% | – |
| ARC-AGI-1 | ARC Prize | 92.5% | 87.5% |
| ARC-AGI-2 | ARC Prize | 72.1% | 61.3% |
| ARC-AGI-3 | ARC Prize | – | 2.1%† |
| CritPt | Artificial Analysis | 13.1% | 17.7% |
| MysteryMechanism | Vals AI | – | 30.6%† |
| Chess Puzzles | Epoch AI Benchmarking Hub | 45.0% | 33.0%† |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 32.0% | 34.0%† |
| LMCA | Epoch AI Benchmarking Hub | 47.1% | 48.5%† |
| DTBench | Epoch AI Benchmarking Hub | 94.7% | 97.3%† |
| ForecastBench | ForecastBench | 64.3%† | 65.5%† |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | 79.3% | – |
| SciCode | SciCode | 53.9% | 55.9% |
| WeirdML | WeirdML | 62.6% | 67.3%† |
| FrontierCode | Cognition | – | 48.0%† |
| Arena Coding | Arena (LMArena) | 1508 | 1510† |
| Arena WebDev | Arena (LMArena) | 1499 | 1617† |
Show all 130 benchmarks
| Metric | Publisher | Gemini 3.5 Flash (high) | Grok 4.6 (medium) |
|---|---|---|---|
| Coding | |||
| LiveBench 2.0 Coding | LiveBench | – | 89.1%† |
| LiveBench 1.0 Coding | LiveBench | 78.2% | 76.8%† |
| LiveCodeBench | Vals AI | 87.6% | 88.2%† |
| IOI | Vals AI | – | 47.6%† |
| SWE-bench (Vals) | Vals AI | 78.8% | 95.6%† |
| Code Migration | Vals AI | 26.8% | 44.6%† |
| ProgramBench | Vals AI | 0.0% | 1.0%† |
| Vibe Code Bench 1-100 | Vals AI | – | 14.8%† |
| Vibe Code Bench v1.1 | Vals AI | 48.7% | 76.2%† |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | 58.1%† | – |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | 36.1% | 67.5% |
| CursorBench | Cursor | – | 36.1% |
| ALE-Bench | Epoch AI Benchmarking Hub | 911.0 | 1508.1† |
| FrontierSWE | Epoch AI Benchmarking Hub | – | 25.3%† |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 6.6% | 13.1% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 78.7% | 84.3% |
| Startrise Frontend | Startrise | – | 72.1%† |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | – | 20.3%† |
| APEX-Agents | Mercor | 27.5%† | 65.3%† |
| Arena Agent | Arena (LMArena) | – | 0.6† |
| LiveBench Long Agentic | LiveBench | – | 40.7%† |
| LiveBench Browser | LiveBench | – | 64.0%† |
| LiveBench Agentic Coding | LiveBench | 49.0% | 57.0%† |
| Terminal-Bench Hard | Artificial Analysis | 40.9% | – |
| τ²-Bench Telecom (AA) | Artificial Analysis | 95.3% | – |
| Terminal-Bench 2.1 (Vals) | Vals AI | 74.2% | 78.3%† |
| MCP Atlas | Scale AI SEAL | 83.6% | – |
| HiL-Bench | Scale AI SEAL | 27.7%† | – |
| τ³-Banking (AA) | Artificial Analysis | 32.2% | 44.3% |
| ITBench SRE (AA) | Artificial Analysis | 40.4% | – |
| Analyst Agent (AA) | Artificial Analysis | 45.0% | 41.3%† |
| APEX-Agents (AA) | Artificial Analysis | 47.0% | – |
| CyberBench | Vals AI | – | 67.7%† |
| Finance Agent v2 | Vals AI | 57.9% | 53.7%† |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 2.5% | 15.8%† |
| Legal Research Bench | Vals AI | 30.8% | 48.1%† |
| SkillsBench | Vals AI | 52.7% | 55.8%† |
| Tax Agent Bench | Vals AI | 21.6% | 33.1%† |
| Terminal-Bench 2.0 (Vals) | Vals AI | 67.4% | – |
| Terminal-Bench 4.0 (Vals) | Vals AI | 6.1% | 17.2%† |
| Terminal-Bench Science | Vals AI | 5.7% | 11.4%† |
| Time Horizon Index: KSP | Vals AI | – | 5.8%† |
| Web Search Index | Vals AI | 37.0% | – |
| GBAEval | Epoch AI Benchmarking Hub | 6.7%† | – |
| Vending-Bench 2 | Andon Labs | 5396.4† | 9047.0† |
| AutomationBench | Artificial Analysis | 42.1% | 63.1% |
| GDP.pdf (AA) | Artificial Analysis | 19.8% | 17.8% |
| EnterpriseOps-Gym | Artificial Analysis | 50.1% | 48.3%† |
| AA-Briefcase v1.1 | Artificial Analysis | 874 | 1531† |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | – | 6.2%† |
| FACTS Search | 64.3%† | – | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 62.8% | 66.0%† |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 26.8% | 31.7%† |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 95.6% | 97.8%† |
| ProofBench | Vals AI | 31.0% | 51.0%† |
| LiveBench Mathematics | LiveBench | 88.2% | 92.6%† |
| AIME 2026 | MathArena | 95.0%† | – |
| HMMT February 2026 | MathArena | 95.5%† | – |
| MathArena Apex | MathArena | 32.3%† | – |
| Arena Maths | Arena (LMArena) | 1497 | 1445† |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 66.2% | 49.3%† |
| LiveBench 2.0 Data Analysis | LiveBench | – | 75.3%† |
| LiveBench Research | LiveBench | – | 84.5%† |
| LiveBench 1.0 Data Analysis | LiveBench | 64.9% | 73.9%† |
| AA-Omniscience | Artificial Analysis | 21.2 | 28 |
| MMLU-Pro | Vals AI | 89.5% | 89.4%† |
| LegalBench | Vals AI | 83.6% | 86.3%† |
| CorpFin | Vals AI | 64.7% | 66.2%† |
| TaxEval | Vals AI | 74.4% | 71.1%† |
| BioMysteryBench | Vals AI | – | 72.2%† |
| Excel Modeling Benchmark | Vals AI | 63.5% | 62.7%† |
| MedCode | Vals AI | 55.8% | 44.7%† |
| MedScribe | Vals AI | 76.6% | 86.5%† |
| Public Benefits Bench v1.1 | Vals AI | 59.5% | 66.8%† |
| DrugDiscoveryBench | Scale AI SEAL | 50.0% | – |
| EBR-bench | Epoch AI Benchmarking Hub | 4.8% | 30.5%† |
| GDP.pdf | Epoch AI Benchmarking Hub | 14.0%† | 16.0%† |
| AA-Omniscience: accuracy | Artificial Analysis | 51.4% | 41.9% |
| AA-Omniscience: non-hallucination | Artificial Analysis | 37.8% | 76.0% |
| Instruction following | |||
| LiveBench Language | LiveBench | 84.6% | 83.7%† |
| LiveBench Instruction Following | LiveBench | 75.6% | 71.9%† |
| IFBench | Artificial Analysis | 76.3% | – |
| Arena Instruction Following | Arena (LMArena) | 1465 | 1452† |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1477 | 1456† |
| EQ-Bench 4 | EQ-Bench | 1081† | – |
| Arena Creative Writing | Arena (LMArena) | 1464 | 1443† |
| Arena Multi-turn | Arena (LMArena) | 1480 | 1448† |
| Arena Longer Queries | Arena (LMArena) | 1482 | 1475† |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1308 | 1263† |
| LiveBench Design | LiveBench | – | 79.8%† |
| MMMU-Pro | Artificial Analysis | 84.3% | – |
| MMMU-Pro (Vals) | Vals AI | 88.3% | – |
| MortgageTax | Vals AI | 68.1% | 64.2%† |
| SAGE | Vals AI | 49.9% | 28.9%† |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 33.6%† | 33.2%† |
| Long context | |||
| AA-LCR | Artificial Analysis | 73.3% | 81.0% |
| Arena Document | Arena (LMArena) | 1453 | 1461† |
| MLCR | Artificial Analysis | 10.0% | 18.3% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 154.9† | 156.6† |
| LiveBench 2.0 | LiveBench | – | 74.9%† |
| LiveBench 1.0 | LiveBench | 74.6% | 78.0%† |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 32.6 | 42.8 |
| Vals RSI Index | Vals AI | – | 25.5%† |
| Vals Science Index | Vals AI | – | 38.1%† |
| Vals Multimodal Index | Vals AI | 62.9% | – |
Data as of 2026-10-25. Best configuration of each model; every score links to its source on the model pages.
What changed
- Gemini 3.5 Flash (high): 1 new result (Harvey's Legal Agent Benchmark v1).
- Gemini 3.5 Flash (high): 1 score revision.
- Grok 4.6 (medium): 1 new result (MLCR).
- Gemini 3.5 Flash (high): BenchLeader Index 62.5 → 63.2.
- Gemini 3.5 Flash (high): 3 new results (GDPval-AA v2.1, AA-Omniscience: accuracy, AA-Omniscience: non-hallucination).
- Grok 4.6 (medium): 15 new results (AA Intelligence Index v4.3.2, GPQA Diamond (AA), Humanity's Last Exam (AA)…).
- Gemini 3.5 Flash (high): 1 score revision.
- Gemini 3.5 Flash (high): 2 score revisions.
- Grok 4.6 (medium): 1 score revision.
- Gemini 3.5 Flash (high): BenchLeader Index 63.0 → 62.5.
Related comparisons
- Gemini 3.5 Flash vs Claude Opus 5.5
- Gemini 3.5 Flash vs Gemini 4 Argon
- Gemini 3.5 Flash vs Muse Wonder Full
- Grok 4.6 vs Claude Opus 5.5
- Grok 4.6 vs Gemini 4 Argon
- Grok 4.6 vs Muse Wonder Full
- Grok 4.6 vs Claude Fable 5.1
- Gemini 3.5 Flash vs GPT-6 Astra
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Gemini 3.5 Flash vs Grok 4.6: questions
- Is Gemini 3.5 Flash better than Grok 4.6?
- Gemini 3.5 Flash (high) and Grok 4.6 (medium) are neck and neck on quality (63.1 vs 64.2, within the ±5.1 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 2026-10-25, so check the category scores for your use.
- Which is cheaper, Gemini 3.5 Flash or Grok 4.6, and will that change?
- Grok 4.6 is cheaper today: $3.00 against $3.38 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Gemini 3.5 Flash better than Grok 4.6 for coding?
- Grok 4.6 scores higher in coding (62 vs 60 on the category index, where 50 is average). On SciCode, the top shared coding benchmark, it is 53.9% for Gemini 3.5 Flash against 55.9% for Grok 4.6.
- Is Gemini 3.5 Flash better than Grok 4.6 for agents and tool use?
- Grok 4.6 scores higher in agentic tasks (74 vs 63 on the category index, where 50 is average). On τ³-Banking (AA), the top shared agents & tools benchmark, it is 32.2% for Gemini 3.5 Flash against 44.3% for Grok 4.6.
- Which is better for long documents?
- Grok 4.6 scores higher on long-context benchmarks (66 vs 62). Gemini 3.5 Flash accepts more context: 1,048,576 against 500,000 tokens for Grok 4.6.
- Which has lower latency?
- Gemini 3.5 Flash (high) answers first: 14.2 s against 25.5 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
- Which is faster, Gemini 3.5 Flash or Grok 4.6?
- Gemini 3.5 Flash streams faster: 222 against 78 output tokens per second.
- Do independent benchmarks agree on Gemini 3.5 Flash vs Grok 4.6?
- Across the 17 benchmarks both have, Gemini 3.5 Flash leads 5, Grok 4.6 leads 12. Independent sources split: Gemini 3.5 Flash leads on Arena Text; Grok 4.6 leads on AA Intelligence Index v4.3.2.
- When was Grok 4.6 released?
- Grok 4.6 was released on 11 Aug 2026.