Gemini 4 Argon vs Inkling 2 Small
Full profiles: Gemini 4 Argon (Google) and Inkling 2 Small (Thinking Machines): every benchmark, price and rank.
Updated · Data from 11 sources
As of 13 Nov 2026, Gemini 4 Argon is ahead of Inkling 2 Small on quality (65.3 vs 57.9 on the BenchLeader Index), but Inkling 2 Small is 8.4× cheaper at $0.475 vs $4.00 per 1M tokens blended. Choose Gemini 4 Argon for knowledge (72 vs 54), reasoning (70 vs 53), and human preference (72 vs 59); choose Inkling 2 Small for agents & tools (61 vs 56) and long context (66 vs 65), where it also streams ~2× faster. Gemini 4 Argon's current price runs until 30 Nov 2026, then $8.00.
Key facts (as of 13 Nov 2026)
- Gemini 4 Argon scores 65.3 on the BenchLeader Index and Inkling 2 Small scores 57.9 (BenchLeader, 13 Nov 2026).
- Gemini 4 Argon costs $4.00 per 1M tokens blended until 30 Nov 2026, then $8.00 (https://ai.google.dev/gemini-api/docs/pricing, checked 20 Oct 2026).
- Inkling 2 Small costs $0.475 per 1M tokens blended (list price, 13 Nov 2026).
- Inkling 2 Small streams ~2× faster than Gemini 4 Argon (Artificial Analysis, 7-day median, 13 Nov 2026).
- Of the 44 benchmarks both models have, Gemini 4 Argon leads 39 (BenchLeader, 13 Nov 2026).
- On Arena WebDev, Gemini 4 Argon is rated 1681 and Inkling 2 Small 1647 (LMArena, as published by 13 Nov 2026).
Gemini 4 Argon benchmarks 
Inkling 2 Small benchmarks Which 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 | Gemini 4 Argon | 67 vs 60 category score; Arena WebDev 1681 vs 1647 | Medium |
| Agents & tool use | Inkling 2 Small | 61 vs 56 category score; Arena Agent 2.5 vs 10.7 | Medium |
| Reasoning | Gemini 4 Argon | 70 vs 53 category score; HLE-Diamond (AA) 56.7% vs 28.7% | High |
| Knowledge | Gemini 4 Argon | 72 vs 54 category score; AA-Omniscience v1.1 46.1 vs 16.1 | High |
| Long documents | Inkling 2 Small | 66 vs 65 category score; AA-LCR v1.1 83.3% vs 79.7% | Medium |
| Chat quality (human preference) | Gemini 4 Argon | 72 vs 59 on Arena votes | High |
| Lowest cost | Inkling 2 Small | $0.475 vs $4.00 per 1M blended; agentic coding session $0.022 vs $0.160; from 1 Dec 2026: $0.475 vs $8.00 | – |
| Fastest first answer | Gemini 4 Argon | 42.7 s (Gemini 4 Argon (high)) vs 63.9 s (Inkling 2 Small) 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 4 Argon (high) leads on quality: 65.3 vs 57.9.
- Gemini 4 Argon (high) is stronger in coding, human preference, knowledge, reasoning.
- Inkling 2 Small is stronger in agents & tools, long context.
- Inkling 2 Small is 8.4× cheaper ($0.475 vs $4.00 per 1M blended) until 30 Nov 2026, then $0.475 vs $8.00 from 1 Dec 2026.
- Inkling 2 Small streams 1.4× faster (90 vs 64 tokens per second).
At a glance
| Spec | Gemini 4 Argon | Inkling 2 Small |
|---|---|---|
| Developer | Thinking Machines | |
| Released | 30 Sep 2026 | 5 Nov 2026 |
| Generally available | 13 Oct 2026 | 5 Nov 2026 |
| Context window (tokens) | 1,048,576 | 1,048,576 |
| Max output (tokens) | 65,536 (up to 1,048,576 with long decode) | – |
| Input types | text, image, video, audio, PDF | – |
| Reasoning effort settings | low, medium, high | – |
| Compared at | Gemini 4 Argon (high) | Inkling 2 Small |
| Open weights | No | No |
| Price in / out / cached input per 1M | $2.00 / $10.00 / $0.100 | $0.300 / $1.00 / – |
| Announced price change | $4.00 / $20.00 / $0.200 from 1 Dec 2026 | None announced |
| API model ID | gemini-4-argon | – |
| Where to use it | Gemini API (AI Studio), Gemini Enterprise Agent Platform, Gemini app | – |
Limits, availability and announced prices from the vendors' own pages: ai.google.dev, ai.google.dev pricing (checked 20 Oct 2026). Everything else is from the day's data.
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 4 Argon (high) | Inkling 2 Small | Leader |
|---|---|---|---|
| Reasoning | 69.5 | 52.8 | Gemini 4 Argon |
| Coding | 66.7 | 60.3 | Gemini 4 Argon |
| Agents & tools | 56.3 | 61.1 | Inkling 2 Small |
| Maths | 69.9 | – | not comparable |
| Knowledge | 71.9 | 53.5 | Gemini 4 Argon |
| Human preference | 72.1 | 58.6 | Gemini 4 Argon |
| Multimodal | – | 64.0 | not comparable |
| Long context | 64.5 | 66.4 | Inkling 2 Small |
On one chart
Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.
Not drawn, measured for fewer than two of these models: Maths, Instruction following, Multimodal.
Where sources disagree
Across the 44 benchmarks both have, Gemini 4 Argon leads 39, Inkling 2 Small leads 5. Every overall index that rates both puts Gemini 4 Argon ahead.
| Overall index | Gemini 4 Argon (high) | Inkling 2 Small | Leader |
|---|---|---|---|
| BenchLeader Index | 65.3 | 57.9 | Gemini 4 Argon |
| AA Intelligence Index v5.0 | 50.4 | 30.6 | Gemini 4 Argon |
| Vals Index | 69.4 | 56.1 | Gemini 4 Argon |
| Arena Text | 1522 | 1412 | Gemini 4 Argon |
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. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.
| Measure | Gemini 4 Argon (high) | Inkling 2 Small |
|---|---|---|
| Blended price per 1M tokens, until 30 Nov 2026 | $4.00 | $0.475 |
| Input / output per 1M | $2.00 / $10.00 | $0.300 / $1.00 |
| Blended price per 1M from 1 Dec 2026 | $8.00 | $0.475 |
| Input / output per 1M from 1 Dec 2026 | $4.00 / $20.00 | $0.300 / $1.00 |
| Cached input per 1M | $0.100 | – |
| Price per task: chat reply | $0.0038 | $0.0004 |
| Price per task: summarize a 30-page report | $0.030 | $0.0042 |
| Price per task: code edit | $0.027 | $0.0033 |
| Price per task: agentic coding session | $0.160 | $0.022 |
| Price per task: structured extraction | $0.0060 | $0.0008 |
| Agentic coding session, input mostly cached | $0.074 | – |
| Cost per benchmark task (AA Index) | $4.97 | $1.18 |
| Output tokens per task (AA Index) | 116k | 150k |
| Of which reasoning | 68k | 88k |
| Time per task (AA Index) | 30 min | 28 min |
| Output speed | 64 tok/s | 90 tok/s |
| Time to first answer | 42.7 s | 63.9 s |
| Time to a complete answer | 122.1 s | 163.6 s |
Per token is not per task: an average task of Artificial Analysis' Intelligence Index cost $4.97 on Gemini 4 Argon (high) and $1.18 on Inkling 2 Small, and Inkling 2 Small wrote 1.3× as many output tokens for the same work.
Each model is compared at its best-ranked reasoning-effort setting. Gemini 4 Argon (high) scores 65.3 and takes 43 s to its first answer token; Gemini 4 Argon scores 64.3 in 5 s, Gemini 4 Argon (medium) scores 61.6 in 5 s. Inkling 2 Small scores 57.9 and takes 64 s to its first answer token. 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 4 Argon (high) | Inkling 2 Small |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 65.3 | 57.9 |
| Agents & tools score | BenchLeader | 56.3 | 61.1 |
| Coding score | BenchLeader | 66.7 | 60.3 |
| Human preference score | BenchLeader | 72.1 | 58.6 |
| Knowledge score | BenchLeader | 71.9 | 53.5 |
| Long context score | BenchLeader | 64.5 | 66.4 |
| Maths score | BenchLeader | 69.9 | – |
| Reasoning score | BenchLeader | 69.5 | 52.8 |
| Multimodal score | BenchLeader | – | 64.0 |
| Blended price $/M | BenchLeader | $4.00 | $0.475 |
| Output speed | BenchLeader | 64 tok/s | 90 tok/s |
| Time to first answer | BenchLeader | 42.7 s | 63.9 s |
| Context window | BenchLeader | 1,048,576 tokens | 1,048,576 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 96.4% | – |
| SimpleBench | SimpleBench | 91.7%† | – |
| Arena Hard Prompts | Arena (LMArena) | 1545 | 1438 |
| LiveBench 2.0 Reasoning | LiveBench | 95.1%† | – |
| Humanity's Last Exam (AA) | Artificial Analysis | 57.1% | 44.1% |
| HLE-Diamond (AA) | Artificial Analysis | 56.7% | 28.7% |
| ARC-AGI-1 | ARC Prize | 98.5% | 96.5%† |
| ARC-AGI-2 | ARC Prize | 91.3% | 81.7%† |
| ARC-AGI-3 | ARC Prize | 97.3% | – |
| CritPt | Artificial Analysis | 27.1% | – |
| CritPt v2 hard subset (AA) | Artificial Analysis | 52.5% | 35.0% |
| MysteryMechanism | Vals AI | 45.5% | 25.2% |
| Chess Puzzles | Epoch AI Benchmarking Hub | 59.0% | – |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 64.0% | – |
| LMCA | Epoch AI Benchmarking Hub | 63.7% | – |
| DTBench | Epoch AI Benchmarking Hub | 97.3% | – |
| Coding | |||
| SciCode | SciCode | 61.8% | 52.3% |
| FrontierCode | Cognition | 47.5%† | – |
| Arena Coding | Arena (LMArena) | 1553 | 1485 |
| Arena WebDev | Arena (LMArena) | 1681 | 1647 |
| LiveBench 2.0 Coding | LiveBench | 85.7%† | – |
| IOI | Vals AI | 100.0% | – |
| Code Migration | Vals AI | 68.2% | – |
| Code Migration v1.1 | Vals AI | 70.8% | – |
| ProgramBench | Vals AI | 2.5% | 0.5% |
Show all 130 benchmarks
| Metric | Publisher | Gemini 4 Argon (high) | Inkling 2 Small |
|---|---|---|---|
| Coding | |||
| SWE-bench Multimodal | Vals AI | 56.0% | 43.1% |
| Vibe Code Bench 1-100 | Vals AI | 21.2% | – |
| Vibe Code Bench v1.1 | Vals AI | 91.9% | 81.5% |
| MirrorCode | Epoch AI Benchmarking Hub | 58.9% | – |
| CursorBench | Cursor | 52.9% | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 2100.1 | – |
| FrontierSWE | Epoch AI Benchmarking Hub | 55.0% | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 57.1% | 13.1% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | – | 81.3% |
| Startrise Frontend | Startrise | 71.0%† | – |
| GameBench 2 Overall | DuelLab GameBench 2 | 81.7† | – |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 89.2 | – |
| Design Arena Website | Design Arena | 1316† | 1335 |
| FrontierCode 1.1 (AA) | Artificial Analysis | 43.5% | – |
| FrontierCode 2.0 (AA) | Artificial Analysis | 25.7% | 25.4% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | – | 13.0%† |
| Remote Labor Index | Scale AI / CAIS | 17.1%† | – |
| APEX-Agents | Mercor | 82.4%† | 46.8% |
| Arena Agent | Arena (LMArena) | 10.7 | 2.5 |
| LiveBench Long Agentic | LiveBench | 58.5%† | – |
| LiveBench Browser | LiveBench | 82.8%† | – |
| Terminal-Bench 2.1 (Vals) | Vals AI | 57.7% | – |
| MCP Atlas | Scale AI SEAL | 90.7%† | 85.4% |
| HiL-Bench | Scale AI SEAL | 62.0%† | – |
| GDPval-AA v2.1, normalised score | Artificial Analysis | 56.2% | – |
| GDPval-AA v2.1 | Artificial Analysis | 1633 | 1630 |
| τ³-Banking (AA) | Artificial Analysis | – | 40.4% |
| Analyst Agent (AA) | Artificial Analysis | 48.8% | 51.3% |
| CUA-bench | Vals AI | 4.8% | – |
| CyberBench | Vals AI | 77.9% | 72.7% |
| Finance Agent v2 | Vals AI | 65.4% | – |
| Finance Agent v2.1 | Vals AI | 67.7% | 67.5% |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | 27.1% | 21.7% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 19.6% | – |
| Legal Research Bench | Vals AI | 54.8% | 46.6% |
| Robo-bench | Vals AI | 0.8% | – |
| SREBench | Vals AI | 44.3% | 1.9% |
| Tax Agent Bench | Vals AI | 49.2% | 38.9% |
| Terminal-Bench 4.0 (Vals) | Vals AI | 57.6% | 12.6% |
| Terminal-Bench Science | Vals AI | 44.3% | 2.9% |
| Time Horizon Index: KSP | Vals AI | 52.5% | – |
| PostTrainBench | Epoch AI Benchmarking Hub | 40.4% | – |
| Vending-Bench 2 | Andon Labs | 13718.2† | – |
| AutomationBench | Artificial Analysis | 77.5% | 45.6% |
| GDP.pdf (AA) | Artificial Analysis | 21.8% | 25.2% |
| Harvey LAB-AA v1.1 | Harvey | 3.9% | – |
| EnterpriseOps-Gym | Artificial Analysis | – | 46.8% |
| AA-Briefcase v1.1 | Artificial Analysis | 1484 | 1527 |
| GDPx | GDPx by SemiAnalysis | 7.3% | – |
| Design Arena Agentic | Design Arena | 1262† | 1374 |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 52.9% | 5.7% |
| FACTS Search | 89.9%† | – | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 86.0% | – |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 78.0% | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 100.0% | – |
| ProofBench | Vals AI | 99.0%† | – |
| Arena Maths | Arena (LMArena) | 1536 | – |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 76.0% | – |
| LiveBench 2.0 Data Analysis | LiveBench | 85.2%† | – |
| LiveBench Research | LiveBench | 86.8%† | – |
| AA-Omniscience v1.1 | Artificial Analysis | 46.1 | 16.1 |
| LegalBench | Vals AI | 88.3% | – |
| PRBench Finance | Scale AI SEAL | 49.9%† | – |
| PRBench Legal | Scale AI SEAL | 49.4%† | – |
| BioMysteryBench | Vals AI | 76.3% | 70.0% |
| Excel Modeling Benchmark | Vals AI | 75.2% | – |
| Excel Modeling Benchmark v1.1 | Vals AI | 74.8% | 69.5% |
| MedCode | Vals AI | 58.8% | – |
| MedScribe | Vals AI | 87.4% | – |
| Public Benefits Bench v1.1 | Vals AI | 69.8% | – |
| DrugDiscoveryBench | Scale AI SEAL | 65.4% | – |
| EBR-bench | Epoch AI Benchmarking Hub | 64.8% | – |
| GDP.pdf | Epoch AI Benchmarking Hub | 28.4% | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 48.3% | 33.8% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 95.7% | 73.3% |
| HealthBench Professional (AA) | OpenAI | 55.0% | – |
| Instruction following | |||
| MultiChallenge | Scale AI SEAL | 78.7%† | – |
| Arena Instruction Following | Arena (LMArena) | 1528 | 1410 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1522 | 1412 |
| EQ-Bench 4 | EQ-Bench | 1236† | – |
| Arena Creative Writing | Arena (LMArena) | 1523 | 1345 |
| Arena Multi-turn | Arena (LMArena) | 1545 | 1415 |
| Arena Longer Queries | Arena (LMArena) | 1544 | 1410 |
| Multimodal | |||
| LiveBench Design | LiveBench | 69.2%† | – |
| MMMU-Pro | Artificial Analysis | 88.3%† | 81.6% |
| SAGE | Vals AI | 89.5% | 88.6% |
| AudioMC | Scale AI SEAL | 58.4% | – |
| AudioMC (text output) | Scale AI SEAL | 58.4% | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 54.5%† | – |
| Code2Video-AA | Artificial Analysis | 1259 | 1020 |
| AA-3D | Artificial Analysis | 1147 | 1027 |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 79.7% | 83.3% |
| MLCR | Artificial Analysis | 13.9% | 20.6% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 166.5† | – |
| LiveBench 2.0 | LiveBench | 77.7%† | – |
| AA Intelligence Index v5.0 | Artificial Analysis | 50.4 | 30.6 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 52.6 | – |
| AA Progress Index v1 | Artificial Analysis | 341.8 | – |
| Vals Index | Vals AI | 69.4 | 56.1 |
| Vals Finance Index | Vals AI | 66.0% | 61.5% |
| Vals RSI Index | Vals AI | 30.6% | – |
| Vals Science Index | Vals AI | 55.4% | 32.7% |
| Vals Router Index | Vals AI | 69.2%† | – |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | – | 16.2% |
| Metadata | |||
| AA Openness Index | Artificial Analysis | – | 38.9 |
Data as of 2026-11-13. Best configuration of each model; every score links to its source on the model pages.
DuelLab GameBench 2, release gb2-20261110-2acac640a228 (https://benchmarks.duellab.org/releases/gb2-20261110-2acac640a228/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.
What changed
- Gemini 4 Argon: list price moves to $4.00 / $20.00 per 1M input / output tokens (announced).
- Inkling 2 Small: BenchLeader Index 58.4 → 57.9.
- Gemini 4 Argon (high): 1 score revision.
- Inkling 2 Small: 2 new results (FORTRESS, Design Arena Agentic).
- Gemini 4 Argon (high): BenchLeader Index 65.9 → 65.3.
- Gemini 4 Argon (high): 1 new result (PostTrainBench); input price moved from $4.00/M to $2.00/M (50% down); output price moved from $20.00/M to $10.00/M (50% down); cost per task moved from $9.95/M to $4.97/M (50% down).
- Inkling 2 Small: 2 new results (MysteryMechanism, Vals Science Index); 1 score revision.
- Inkling 2 Small: 3 new results (SciCode, APEX-Agents, MLCR).
- Gemini 4 Argon (high): 4 score revisions.
- Inkling 2 Small: 8 new results (Arena Text, Arena Hard Prompts, Arena Coding…); 2 score revisions.
Related comparisons
- Gemini 4 Argon vs Claude Fable 5.2
- Gemini 4 Argon vs Gemini 4 Carbon
- Gemini 4 Argon vs Claude Opus 5.5
- Inkling 2 Small vs Gemini 4 Carbon
- Inkling 2 Small vs Claude Opus 5.5
- Inkling 2 Small vs GPT-6 Astra
- Inkling 2 Small vs Claude Fable 5.1
- Gemini 4 Argon vs Claude Fable 5.1
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Gemini 4 Argon vs Inkling 2 Small: questions
- Is Gemini 4 Argon better than Inkling 2 Small?
- Gemini 4 Argon (high) leads on quality: 65.3 vs 57.9. The BenchLeader Index combines every independent quality benchmark; Gemini 4 Argon (high) is ahead overall as of 13 Nov 2026, but check the category scores for your use.
- Gemini 4 Argon vs Inkling 2 Small: which is cheaper, and will that change?
- Inkling 2 Small is cheaper today: $0.475 against $4.00 per million tokens, blended at three input tokens per output token. Gemini 4 Argon's current price runs until 30 Nov 2026, then $8.00.
- Is Gemini 4 Argon or Inkling 2 Small better for coding?
- Gemini 4 Argon scores higher in coding (67 vs 60 on the category index, where 50 is average). On Arena WebDev, the top shared coding benchmark, it is 1681 for Gemini 4 Argon against 1647 for Inkling 2 Small.
- Which is faster, Gemini 4 Argon or Inkling 2 Small?
- Inkling 2 Small streams ~2× faster (90 vs 64 output tokens per second, Artificial Analysis 7-day median).
- Is Gemini 4 Argon or Inkling 2 Small better for agents and tool use?
- Inkling 2 Small scores higher in agentic tasks (61 vs 56 on the category index, where 50 is average). On Arena Agent, the top shared agents & tools benchmark, it is 10.7 for Gemini 4 Argon against 2.5 for Inkling 2 Small.
More questions (4)
- Which is better for long documents?
- Inkling 2 Small scores higher on long-context benchmarks (66 vs 65). Both accept 1,048,576 tokens of context.
- Which has lower latency?
- Gemini 4 Argon answers first: 42.7 s against 63.9 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.
- Do independent benchmarks agree on Gemini 4 Argon vs Inkling 2 Small?
- Across the 44 benchmarks both have, Gemini 4 Argon leads 39, Inkling 2 Small leads 5. Every overall index that rates both puts Gemini 4 Argon ahead.
- When was Inkling 2 Small released?
- Inkling 2 Small was released on 5 Nov 2026.