Claude Fable 5.1 vs Gemini 4 Carbon

Claude Fable 5.1 is served by invitation. Its figures below are the live rates, but calls need a workspace grant first; ask us for access before you build on this comparison.

vs

Which one, when

Both run near a million tokens of context (1000000 for claude-fable-5-1, 1048576 for gemini-4-carbon), but the rate cards are far apart: gemini-4-carbon is 5x cheaper on input ($2 against $10) and output ($10 against $50), and 2.5x cheaper on cache reads ($0.1 against $0.25). Pick gemini-4-carbon when cost dominates or the input includes audio or video, which claude-fable-5-1 does not take. Pick claude-fable-5-1 for text and image work that needs replies up to 128000 tokens, about 2x the 65536 on gemini-4-carbon.

Benchmarks

LeadsAbove averageNo peer higherClaude Fable 5.1019 / 323 / 32Gemini 4 Carbon1112 / 1210 / 12

11 measured on both.

Claude Fable 5.1 Gemini 4 Carbon other models measured peer average ★ no peer scored higher
FrontierSWE 2 Proximal leaderboard, Carbon report
56.3%
61.2%
OSWorld 2.1 82 offline tasks, partial score. Gemini: best of 3. Peers: published evaluations
76.8%
81.5%
HealthBench Professional
58.1%
N/A
AutomationBench (Zapier) Zapier private set
31.4%
no peer scored higher 53.9%
GraphWalks BFS (F1) 256k to 1M tokens, 200 tasks, Google evaluation
65%
no peer scored higher 96.2%
Blueprint-Bench 2 Similarity score, Andon Labs leaderboard
41.9%
no peer scored higher 58.6%
Vending-Bench 2 (USD) Andon Labs leaderboard, Carbon report · 5422-16484
5422
no peer scored higher 16484
Chartography No tools, Carbon report
46.2%
no peer scored higher 72.5%

Vendor-published: Alibaba (Qwen) Alibaba (Qwen) Alibaba (Qwen) Alibaba (Qwen) Alibaba (Qwen) Alibaba (Qwen) Alibaba (Qwen) Anthropic Anthropic Anthropic Anthropic ByteDance DeepSeek DeepSeek Google Google Google Google MiniMax Moonshot OpenAI OpenAI OpenAI OpenAI OpenAI OpenAI OpenAI Tencent Z.ai Z.ai Z.ai Z.ai

Pricing

Claude Fable 5.1 Gemini 4 Carbon Δ
Input / 1M tokens $10 $2 5×
Output / 1M tokens $50 $10 5×
Cache read / 1M tokens $0.25 $0.1 2.5×
Cache write 1.25x (5m) / 2x (1h) - -

Rates from the live catalogue at build time; each model page carries the current rate card.

Where they sit · input price per 1M tokens across all 86 chat models on this billing unit (log scale)

Capabilities

Claude Fable 5.1 Gemini 4 Carbon
Tool calling yes yes
Thinking control always on always on
Structured output yes yes
Prompt caching explicit (you mark the prefix) implicit (automatic)
Cache lifetime 5m default, 1h option not published
Minimum cached prefix 512 tokens 4,096 tokens

Specs

Claude Fable 5.1 Gemini 4 Carbon
Input modalities text image text image audio video
Output modalities text text
Released 2026-09-01 2026-11-10
Knowledge cutoff 2026-06 2026-03
Context window 1,000,000 1,048,576
Max output 128,000 65,536
Thinking parameter output_config.effort (thinking is adaptive-only and always on) thinking_level
Accepted values
effort
  • low
  • medium
  • high
  • xhigh
  • max
thinking_level
  • low
  • medium
  • high

"minimal" and thinking_budget return 400

Default

thinking always on (adaptive)

effort
  • high
medium

Specs are transcribed from each vendor’s documentation; a row a vendor does not publish is left out rather than inferred. Full sources: Claude Fable 5.1 · Gemini 4 Carbon

One prompt, both models, measured through the gateway

PROMPT Explain to a CFO, in exactly three sentences, why our LLM bill doubled after we enabled reasoning. CHECK exactly 3 sentences

Claude Fable 5.1 passed · 3 sentences

When we enabled reasoning, the model started generating a long chain of internal "thinking" tokens before producing each answer, and we're billed for those tokens at the same output rate even though users never see them. In practice, that hidden reasoning often runs several times longer than the visible response, so the total tokens per request roughly doubled while the number of requests stayed flat. The upside is measurably better accuracy on complex tasks, so the right conversation is whether that quality gain justifies the cost—and if not, we can restrict reasoning to the workflows that actually need it rather than turning it on globally.

out 186 tok latency 3.2 s

Gemini 4 Carbon passed · 3 sentences

Enabling reasoning instructs the AI to generate a lengthy, behind-the-scenes "thought process" to verify its logic before delivering a final answer. Because our vendor bills us based on total output volume, we are charged for all of that invisible scratchpad work in addition to the visible response, effectively doubling the compute consumed per query. Moving forward, we can control this cost by restricting reasoning strictly to complex, high-stakes workflows where higher accuracy justifies the premium, while routing routine tasks back to standard, lower-cost models.

out 908 tok (+799 thinking) latency 18.0 s

Instruction following (exactly three sentences, so countable), audience shaping (CFO register), and the hidden-thinking billing gap the token meters below expose.

PROMPT This function is supposed to return the longest run of consecutive equal items, but callers report it is off by one on single-element inputs and misses runs that end at the last index. Fix it and explain the bug in one paragraph. CHECK fix passes tests

Claude Fable 5.1 passed · 8/8 cases

```python def longest_run(xs): if not xs: return 0 best = cur = 1 for i in range(1, len(xs)): cur = cur + 1 if xs[i] == xs[i - 1] else 1 best = max(best, cur) return best ``` The bug is that `cur` counts *matches between adjacent elements* rather than *elements in the run*: it starts at 0 and only increments when `xs[i] == xs[i-1]`, so a run of length k produces a count of k−1. That's why a single element returns 0 instead of 1, and why every other run also comes out one short — the "misses runs at the end" reports are the same off-by-one seen on inputs like `[1, 2, 2]`, where the trailing run of length 2 is reported as 1 (the original code does update `best` inside the loop, so end-of-list runs aren't actually skipped, just undercounted). […]

out 882 tok (+507 thinking) latency 3.4 s

Gemini 4 Carbon passed · 8/8 cases

```python def longest_run(xs): if not xs: return 0 best = 1 cur = 1 for i in range(1, len(xs)): if xs[i] == xs[i - 1]: cur += 1 best = max(best, cur) else: cur = 1 return best ``` The original function counts the number of matching adjacent *transitions* rather than the number of *elements* in a run because `best` and `cur` are initialized and reset to `0` instead of `1`. Because a single item on its own already forms a run of length `1`, starting at `0` causes every non-empty result to be short by one: single-element lists (where the loop never executes) and lists with no duplicates return `0` instead of `1`, and any run of $k$ equal items—including one ending at the last index—only increments `cur` $k - 1$ times, making it appear as though the final element of the run was missed. […]

out 1411 tok (+1131 thinking) latency 32.7 s

Whether the fix is actually correct (runnable), explanation density, and token efficiency on a bounded task.

PROMPT Extract a JSON object with fields {company, quarter, revenue_usd_m, growth_pct, guidance} from this text. Use null for anything not stated; add no other fields. CHECK valid JSON, schema exact

Claude Fable 5.1 passed · 5/5 fields, guidance null

```json { "company": "Meridian Robotics", "quarter": "Q3 (September quarter)", "revenue_usd_m": 412, "growth_pct": 11, "guidance": null } ``` Notes on judgment calls (not part of the object): - **revenue_usd_m**: The text says "412 for the topline" but never states the currency or that the figure is in millions; it's also on a constant-currency basis and prefaced with "call it," so it's approximate. If you want strict literalism, this should be `null`. - **growth_pct**: "Eleven and change" means slightly above 11%; the exact figure isn't given. - **quarter**: "September quarter" is Q3 on a calendar basis, consistent with the reference to a pending Q4 close. - **guidance**: Explicitly withheld, so `null`.

out 633 tok (+349 thinking) latency 3.8 s

Gemini 4 Carbon passed · 5/5 fields, guidance null

```json { "company": "Meridian Robotics", "quarter": "September quarter", "revenue_usd_m": null, "growth_pct": null, "guidance": null } ```

out 3338 tok (+3281 thinking) latency 61.6 s

Schema obedience (no invented fields), hallucination pressure (guidance is explicitly withheld), and structured-output path differences.

PROMPT Write a 120-word product announcement for an API gateway feature that caches prompts across providers. Forbidden words: "seamless", "unlock", "game-changer", "revolutionize", "empower". Exactly one sentence must be a question. CHECK 120 words, 0 banned words

Claude Fable 5.1 passed · 121 words, 0 banned, 1 question

**Introducing Cross-Provider Prompt Caching for the Gateway.** Starting today, your API gateway can cache prompts once and reuse them across OpenAI, Anthropic, Google, and any other provider you route to. Long system instructions, retrieval context, and few-shot examples are stored at the gateway layer, so switching models no longer means paying to resend the same tokens. Why send the same 8,000-token preamble to three different vendors when you could send it once? Our cache normalizes prompt segments, fingerprints them, and serves matching prefixes from memory before the request ever reaches an upstream API. Early customers report token costs falling by 30 to 60 percent on high-volume workloads, with median latency dropping alongside. Available today on all Team and Enterprise plans.

out 1722 tok (+1463 thinking) latency 2.9 s

Gemini 4 Carbon passed · 120 words, 0 banned, 1 question

Are your engineering teams paying multiple AI vendors to generate identical responses? Today, we are introducing Universal Prompt Caching for our flagship API Gateway. When applications route traffic across providers like OpenAI, Anthropic, and Mistral, duplicate queries drain budgets and degrade user experiences. Our new unified caching layer intercepts incoming prompts directly at the network edge. If an identical request was previously fulfilled by any connected model, the gateway delivers the stored response in milliseconds, completely bypassing external API calls. This practical upgrade cuts monthly token expenses substantially while accelerating overall application speed. Best of all, deployment requires zero complex code changes. Simply activate this capability inside your gateway configuration dashboard today to start reducing latency and infrastructure costs immediately.

out 2620 tok (+2480 thinking) latency 31.1 s

Constraint obedience (word budget, banned-word list, the single question), style fingerprint, and length control.

Switch between them with one line

Both ids are in every tab below; the highlighted pair of lines is the only edit. Same endpoint, same key, same request shape.

from openai import OpenAI

client = OpenAI(
    base_url="https://synthorai.io/v1",
    api_key="sk-syn-...",
)

resp = client.chat.completions.create(
    model="claude-fable-5-1",
    # model="gemini-4-carbon",  # uncomment this line, comment the one above
    messages=[{"role": "user", "content": "Summarize this diff"}],
    reasoning_effort="medium",
)
print(resp.choices[0].message.content)

Get your API key →

FAQ

Which is cheaper, Claude Fable 5.1 or Gemini 4 Carbon?

Gemini 4 Carbon is cheaper on the "Input / 1M tokens" row ($2 vs $10, 5.0× apart). Other rows may point the other way; the table above carries the full rate card, and real cost depends on your mix.

Can I A/B test Claude Fable 5.1 against Gemini 4 Carbon without two integrations?

Yes. Both are served through the same OpenAI-compatible endpoint with one API key. Switching is a one-line change to the model id, so you can route a fraction of traffic to each and compare bills directly.

Do Claude Fable 5.1 and Gemini 4 Carbon support prompt caching?

Yes. Both bill cache reads below their input rate, so warm-prefix workloads cost less than the list rates suggest. The exact cache-read rows are in the pricing table above.

Related comparisons

From our measured studies