Compare/Sarvam 105B (high) vs Qwen3.5 4B (Reasoning)

Sarvam 105B (high)vsQwen3.5 4B (Reasoning)

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

Sarvam

Sarvam 105B (high)

Input
$0.042/M
Output
$0.17/M
Speed
155 tok/s
TTFT
1.20s
Alibaba

Qwen3.5 4B (Reasoning)

Input
$0.03/M
Output
$0.15/M
Speed
32 tok/s
TTFT
0.50s

Winner by Category

Cheaper
Qwen3.5 4B (Reasoning)
Faster (tok/s)
Sarvam 105B (high)
Lower Latency
Qwen3.5 4B (Reasoning)
Benchmarks (2-5)
Qwen3.5 4B (Reasoning)

Pricing Comparison

MetricSarvam 105B (high)Qwen3.5 4B (Reasoning)
Input ($/M tokens)$0.042$0.03
Output ($/M tokens)$0.17$0.15
Cost for 1M input + 100K output tokens:
Sarvam 105B (high)$0.06
Qwen3.5 4B (Reasoning)$0.04

Speed Comparison

Output Speed (tokens/s) — higher is better
Sarvam 105B (high)
155 tok/s
Qwen3.5 4B (Reasoning)
32 tok/s
Time to First Token (seconds) — lower is better
Sarvam 105B (high)
1.20s
Qwen3.5 4B (Reasoning)
0.50s

Benchmark Comparison

Data from Artificial Analysis API — 12 benchmarks

Intelligence Index
11.920.1
Coding Index
22.6
Math Index
GPQA Diamond
73.8%77.1%
MMLU-Pro
LiveCodeBench
AIME 2025
MATH-500
Humanity's Last Exam
10.1%7.8%
SciCode
26.4%16.1%
IFBench
34.4%52.0%
TerminalBench
1.5%18.2%
Sarvam 105B (high)2 wins
5 winsQwen3.5 4B (Reasoning)

Frequently Asked Questions

Which is cheaper, Sarvam 105B (high) or Qwen3.5 4B (Reasoning)?

Qwen3.5 4B (Reasoning) is cheaper overall. Its blended price (3:1 input/output ratio) is $0.06/M tokens vs $0.07/M for Sarvam 105B (high).

Which model performs better on benchmarks?

Qwen3.5 4B (Reasoning) wins 5 out of 12 benchmarks compared to 2 for Sarvam 105B (high). See the detailed benchmark chart above for per-category results.

Which is faster for real-time applications?

Sarvam 105B (high) generates tokens faster at 155 tok/s vs 32 tok/s. However, Qwen3.5 4B (Reasoning) has lower time-to-first-token (0.50s vs 1.20s).

When should I use Sarvam 105B (high) vs Qwen3.5 4B (Reasoning)?

Choose based on your priorities: Qwen3.5 4B (Reasoning) for lower cost, Qwen3.5 4B (Reasoning) for stronger benchmark performance, and Sarvam 105B (high) for faster generation. For latency-sensitive apps, check the TTFT comparison above.