Compare / Claude Sonnet 4.5 vs Gemma 4 26B A4B

Claude Sonnet 4.5 vs Gemma 4 26B A4B

Which is cheaper, which is more capable, and which does better at each kind of work, from 23 results on 4 sources that measured both models.

AnthropicClaude Sonnet 4.5 Released 29 Sep 2025 · $3.00 in · $15.00 out
GoogleGemma 4 26B A4B Released 2 Apr 2026 · $0.090 in · $0.30 out
Claude Sonnet 4.5 wins 163 too close to callGemma 4 26B A4B wins 4

Of the 23 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.

Results as of 16 November 2026 (catalogue release 2026-11-16-49bdfa045dee). See also Anthropic vs Google, every model from both labs.

Claude Sonnet 4.5 or Gemma 4 26B A4B? The short answer

Pick Claude Sonnet 4.5 if quality matters most: it scores higher (Intelligence Index 20.7 against 16.7). Pick Gemma 4 26B A4B if cost matters more: it costs 42 times less per token.

  • Gemma 4 26B A4B is cheaper: $0.14 against $6.00 per million tokens, 42 times less.
  • Claude Sonnet 4.5 scores higher on the Artificial Analysis Intelligence Index: 20.7 against 16.7.
  • Gemma 4 26B A4B is newer: released 2 Apr 2026, against 29 Sep 2025 for Claude Sonnet 4.5.
  • On third-party benchmarks, on the results both models have: Claude Sonnet 4.5 does better for agents and tool use, reasoning and knowledge and writing; Gemma 4 26B A4B for sticking to the facts.

Head to head

20.7
Intelligence Index
16.7
$6.00
Price per million tokens, blendedlower is better
$0.14
1,000,000
Context window, tokens
262,144

Every fact side by side

Claude Sonnet 4.5Gemma 4 26B A4B
Price per million tokens$3.00 in · $15.00 out$0.090 in · $0.30 out
Blended (3 in : 1 out)$6.00$0.14
Cheapest host, blended$6.00 (Amazon Bedrock)$0.068 (Darkbloom)
Intelligence Index20.716.7
Spring Prompt overall––
Output speed, tokens a second–82
Context window1,000,000 tokens262,144 tokens
Released29 Sep 20252 Apr 2026 (newer)
WeightsClosedOpen
Developer based inthe USthe US

Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.

Which is better for what

Each model's rank among the models on sale today, for each kind of work, from our own benchmarks and licensed sources. Further right is better; the better rank is in green.

Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 103 of 159 105 of 159
Professional workReal tasks from banking, consulting and law, business documents and freelance projects 40 of 51 not measured
Reasoning and knowledgeHard questions across science, maths and general knowledge 104 of 187 110 of 187
WritingWhat people prefer in blind comparisons, and judged writing quality 35 of 169 82 of 169
Sticking to the factsSummarising without inventing things, and factual answers 94 of 139 43 of 139

From our research

Every shared result

Each model's best result on every measure both have, grouped by benchmark, with the setting that produced it. Under each measure, every model that benchmark has measured is a grey tick, best to the right, so you can see where the two sit in the field. The headline measures of each benchmark come first; the rest are under "Show all". Green marks the better value; where a source reports ranges that overlap, neither is marked.

MeasureClaude Sonnet 4.5Gemma 4 26B A4B
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results
Overall · TextArena rating, higher is better 56 · 87 of 387 measured 1,457high-32k reasoning 1,437
Overall · Text factualityArena rating, higher is better 61 · 71 of 163 measured 1,449 1,445
Business, management and finance · TextArena rating, higher is better 51 · 78 of 380 measured 1,461 1,444
Artificial Analysis reported by Artificial Analysis · 3 results
Artificial Analysis Intelligence Indexindex score, higher is better 107 · 123 of 249 measured 20.7thinking reasoning 16.7
GPQA Diamond% of questions, higher is better 95 · 113 of 232 measured 83.4%thinking reasoning 79.2%
Humanity's Last Exam% of questions, higher is better 124 · 119 of 250 measured 17.8%thinking reasoning 19.3%
UGI Leaderboard reported by UGI Leaderboard · 3 results
Writing score · UGIscore out of 100, higher is better 23 · 92 of 230 measured 66.9no reasoning 43.8channel-thought-prefill reasoning
Requested-length error · UGI% off the requested word count, lower is better 15 · 25 of 230 measured 5.0%no reasoning 8.0%
Style adherence · UGIscore from 0 to 1, higher is better 8 · 93 of 230 measured 0.41thinking reasoning 0.35channel-thought-prefill reasoning
Vectara Hallucination Leaderboard reported by Vectara Hallucination Leaderboard · 2 results
Hallucination rate · Vectara% of summaries, lower is better 77 · 13 of 104 measured 12.0% 5.2%
Answer rate · Vectara% of documents, higher is better 86 · 31 of 104 measured 95.6% 99.8%
Show all 23 results (12 more)
MeasureClaude Sonnet 4.5Gemma 4 26B A4B
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 4 results
Creative writing · TextArena rating, higher is better 27 · 96 of 385 measured 1,454 1,400
Expert prompts · TextArena rating, higher is better 43 · 71 of 338 measured 1,498high-32k reasoning 1,475
Instruction following · TextArena rating, higher is better 38 · 71 of 387 measured 1,464high-32k reasoning 1,437
Writing, literature and language · TextArena rating, higher is better 38 · 97 of 386 measured 1,454 1,406
Artificial Analysis reported by Artificial Analysis · 8 results
Terminal-Bench 2.1% of tasks, higher is better 54 · 75 of 131 measured 55.8%thinking reasoning 39.0%
τ-bench banking% of tasks, higher is better 40 · 70 of 121 measured 24.5%thinking reasoning 12.0%
Artificial Analysis Coding Indexindex score, higher is better 54 · 74 of 130 measured 52.1thinking reasoning 39.3
IFBench% of instructions, higher is better 85 · 40 of 195 measured 57.3%thinking reasoning 72.4%
Long-context reasoning (AA-LCR)% of questions, higher is better 102 · 131 of 241 measured 72.3%thinking reasoning 65.7%
SciCode% of problems, higher is better 71 · 86 of 121 measured 45.7%thinking reasoning 40.0%
Terminal-Bench Hard% of tasks, higher is better 47 · 78 of 190 measured 35.6%thinking reasoning 25.0%no reasoning
τ²-bench telecom% of tasks, higher is better 73 · 110 of 193 measured 78.1%thinking reasoning 43.6%

Quick answers

Is Claude Sonnet 4.5 better than Gemma 4 26B A4B?

Pick Claude Sonnet 4.5 if quality matters most: it scores higher (Intelligence Index 20.7 against 16.7). Pick Gemma 4 26B A4B if cost matters more: it costs 42 times less per token.

Which is cheaper, Claude Sonnet 4.5 or Gemma 4 26B A4B?

Gemma 4 26B A4B is cheaper: $0.14 against $6.00 per million tokens, 42 times less.

Which is better for agents and tool use, Claude Sonnet 4.5 or Gemma 4 26B A4B?

Claude Sonnet 4.5 ranks 103 of 159 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 105 for Gemma 4 26B A4B.

Which is better for reasoning and knowledge, Claude Sonnet 4.5 or Gemma 4 26B A4B?

Claude Sonnet 4.5 ranks 104 of 187 models on sale for hard questions across science, maths and general knowledge, against 110 for Gemma 4 26B A4B.

Which is better for writing, Claude Sonnet 4.5 or Gemma 4 26B A4B?

Claude Sonnet 4.5 ranks 35 of 169 models on sale for what people prefer in blind comparisons, and judged writing quality, against 82 for Gemma 4 26B A4B.

Which is better for sticking to the facts, Claude Sonnet 4.5 or Gemma 4 26B A4B?

Gemma 4 26B A4B ranks 43 of 139 models on sale for summarising without inventing things, and factual answers, against 94 for Claude Sonnet 4.5.

Which has the bigger context window, Claude Sonnet 4.5 or Gemma 4 26B A4B?

Claude Sonnet 4.5: 1,000,000 tokens, against 262,144 for Gemma 4 26B A4B.

Run Claude Sonnet 4.5 and Gemma 4 26B A4B on your own prompt.

Benchmarks aren't your task. Test both side by side on your prompt, with every answer checked against your rules and the cost of each one.