Compare/Celeris-1 vs Mercury 2

Celeris-1vsMercury 2

Side-by-side comparison of pricing, 12 benchmarks, and generation speed.

Celeris

Celeris-1

Input
$0.2/M
Output
$0.7/M
Speed
1484 tok/s
TTFT
0.58s
Inception

Mercury 2

Input
$0.25/M
Output
$0.75/M
Speed
835 tok/s
TTFT
3.93s

Winner by Category

Cheaper
Celeris-1
Faster (tok/s)
Celeris-1
Lower Latency
Celeris-1
Benchmarks (0-2)
Mercury 2

Pricing Comparison

MetricCeleris-1Mercury 2
Input ($/M tokens)$0.2$0.25
Output ($/M tokens)$0.7$0.75
Cost for 1M input + 100K output tokens:
Celeris-1$0.27
Mercury 2$0.33

Speed Comparison

Output Speed (tokens/s) — higher is better
Celeris-1
1484 tok/s
Mercury 2
835 tok/s
Time to First Token (seconds) — lower is better
Celeris-1
0.58s
Mercury 2
3.93s

Editorial Analysis

Verdict. Mercury 2 takes the aggregate benchmark matchup 2–0 across 2 categories. Real workloads usually care about a handful of specific tasks — see the per-benchmark table above.

Pricing. Both models sit in the budget bracket for output-token pricing. At 0.9× the per-million-token cost, Celeris-1 is meaningfully cheaper if your traffic is output-heavy (long completions, document generation, agent loops). Celeris-1 makes more sense when output volume is low and absolute reasoning quality justifies the premium.

Strengths. Celeris-1 is strongest on Coding Index (14.4), Intelligence Index (12.4). Mercury 2 leads on Coding Index (31.1), Intelligence Index (21.9).

Speed. On throughput, Celeris-1 generates tokens at 1484 tok/s versus 835 tok/s — about 44% faster. On time-to-first-token, Celeris-1 responds in 580ms vs 3930ms, which matters most for chat-style UIs.

Provider. Celeris and Inception sell to overlapping but distinct developer audiences: Celeris tends to ship frontier reasoning models with premium positioning, while Inception often prices more aggressively. Your existing vendor relationships, billing, and SLA preferences may matter as much as the raw numbers above.

Workload cost. Workload scenarios (per million requests at 30M input + 15M output tokens): Celeris-1 costs $16.50 ($198/year); Mercury 2 costs $18.75 ($225/year). At a smaller 5M-input/2M-output scale (single-developer tool or prototype): Celeris-1 ≈ $2.40/run, Mercury 2 ≈ $2.75/run. At agent/realtime scale (200M input / 100M output per million requests): Celeris-1 ≈ $110/run, Mercury 2 ≈ $125/run. Celeris-1 becomes more attractive at higher volume — the absolute per-token pricing difference compounds when you ship at scale.

Recommendation. Both models have legitimate use cases — the right answer depends on whether you are optimizing for benchmark ceiling, latency, or unit cost. Start with the cheaper / faster model, evaluate against your specific task, and only switch if the upgrade shows a meaningful lift.

Head-to-head deltas

  • On throughput, Celeris-1 is 1.78× faster (1484 tok/s vs 835 tok/s). For streaming chat or real-time agents this alone often flips the recommendation.
  • Time-to-first-token differs by 6.8× — Celeris-1 responds in 580ms vs 3930ms. For interactive chat UIs this can matter more than raw benchmark wins.

Benchmark Comparison

Data from Artificial Analysis API — 12 benchmarks

Intelligence Index
12.421.9
Coding Index
14.431.1
Math Index
GPQA Diamond
MMLU-Pro
LiveCodeBench
AIME 2025
MATH-500
Humanity's Last Exam
SciCode
IFBench
TerminalBench
Celeris-10 wins
2 winsMercury 2

Frequently Asked Questions

Which is cheaper, Celeris-1 or Mercury 2?

Celeris-1 is cheaper overall. Its blended price (3:1 input/output ratio) is $0.33/M tokens vs $0.38/M for Mercury 2.

Which model performs better on benchmarks?

Mercury 2 wins 2 out of 12 benchmarks compared to 0 for Celeris-1. See the detailed benchmark chart above for per-category results.

Which is faster for real-time applications?

Celeris-1 generates tokens faster at 1484 tok/s vs 835 tok/s. Celeris-1 also has lower time-to-first-token (0.58s vs 3.93s).

When should I use Celeris-1 vs Mercury 2?

Choose based on your priorities: Celeris-1 for lower cost, Mercury 2 for stronger benchmark performance, and Celeris-1 for faster generation. For latency-sensitive apps, check the TTFT comparison above.