Towards AITowards AIToneBench

Model Comparison

Claude Opus 5 vs Claude Haiku 5.5

Claude Opus 5 (max effort) leads Claude Haiku 5.5 (max effort) by 321 Elo, and their 95% ranges sit 236 Elo apart. Even Claude Opus 5's lowest-rated configuration, Claude Opus 5 (high), ranks above Claude Haiku 5.5 (max effort). Among 84 models, each ranked by its best configuration, Claude Opus 5 is #5 and Claude Haiku 5.5 is #19. Cost per script: $0.736 for Claude Opus 5 (max effort) and $0.103 for Claude Haiku 5.5 (max effort).

Last updated , from board data generated at 20:15 UTC.

Claude Opus 5
2287 Writing Elo of Claude Opus 5 (max effort), its best configuration#5 of 84 models · 83.3/100 Writing Score · $0.736 per script2 configurations on the board
Claude Haiku 5.5
1966 Writing Elo of Claude Haiku 5.5 (max effort), its best configuration#19 of 84 models · 80.8/100 Writing Score · $0.103 per script5 configurations on the board

Picks and Verdict

PickClaude Opus 5

Claude Opus 5 (high)

Pick Claude Opus 5 (high) when the voice has to carry the script, since it already beats every Haiku setting for $0.472 a script.

PickClaude Haiku 5.5

Claude Haiku 5.5 (xhigh)

Pick Claude Haiku 5.5 (xhigh) for cheap volume drafts at $0.029 a script, since its range overlaps max effort's, which costs about 3.6 times as much and takes about 13 minutes.

Claude Opus 5 (high), the cheaper of its two settings, already clears Claude Haiku 5.5 (max effort) by 218 Elo, so quality isn't a close call. Every Opus 5 setting rates above every Haiku setting, and none of their 95% ranges touch. Opus pulls away on structure and flow: at max effort it leads Haiku's best by about seven points on YouTube best practices, five on continuity and four on the hook. To be fair to Haiku, it beats Opus on substance, length and slop, so the facts and the format come out fine. The Haiku ladder is the part I'd look at twice. Max effort costs $0.103 and takes about 13 minutes, xhigh costs $0.029 and takes under four. Max sits 71 Elo above xhigh, but their ranges overlap, and max costs about 3.6 times as much and takes more than three times as long. Below xhigh it drops fast: high sits 209 Elo lower. If the voice has to come from the model, I'd pay $0.472 for Opus at high and keep max effort, 103 Elo higher with steadier drafts and a range that just clears high's, for a script I record. If cost and turnaround rule, Haiku at xhigh is the one to run.

Disclosure: both models come from the same lab as one of the four judges, Claude Opus 5.5. The other judges sit outside that lab, so neither is scored by its own lab alone (how the panel works). The judge analysis measures how much each judge favors its own lab.

How to Read the Writing Score, Writing Elo and 95% Ranges

Writing Score and Writing Elo. The Writing Score (0 to 100) is the weighted average of the nine metric scores: seven from the judges, Length Adherence from code and the Slop Score half from each. The Writing Elo is a rating from pairwise comparisons: on each script, each configuration's average across its drafts is compared with every other configuration's, and gaps within the draft-to-draft variation count as draws. Elo rewards winning often, so the two can put configurations in a different order.

Elo ranges. Each 95% range shows how far a configuration's Elo could move if the drafts had come out slightly differently. It comes from resampling every configuration's recorded draft scores on each reference script and ranking again. Comparing two separate ranges does not test the Elo gap between them: ranges far apart are strong evidence for the order, and overlapping ranges do not mean a tie. The reference scripts stay fixed, so the ranges say nothing about new scripts. The methodology page has the full description.

All Thinking Levels of Both Models

Every ranked configuration of Claude Opus 5 and Claude Haiku 5.5, scored on the same 10 scripts by the same four judges.

How to Read the Chart
  • Each row is one configuration: the model at one thinking level. The dot is its Writing Elo, a rating from pairwise comparisons of the scores on each script, and the bar is its 95% range: how far that Elo could move if the drafts had come out slightly differently.
  • Each bar describes one configuration. Overlap alone does not establish a tie or decide their order. Comparing these ranges does not directly test the difference between a pair.
  • The table under the chart lists the same configurations in the same order and adds the Writing Score (the 0 to 100 consensus of the judges and code), cost, time and configuration rank (the place among all configurations on the board).
Claude Opus 5Claude Haiku 5.5
ConfigurationWriting Elo95% RangeWriting ScoreCost / ScriptTime / ScriptConfiguration Rank
Claude Opus 5 (max effort)22872242 to 233083.3$0.7365.5 min#11
Claude Opus 5 (high)21842137 to 222382.6$0.4723.5 min#16
Claude Haiku 5.5 (max effort)19661925 to 200680.8$0.10312.6 min#35
Claude Haiku 5.5 (xhigh)18951854 to 193780.0$0.0293.6 min#50
Claude Haiku 5.5 (high)16861647 to 173378.0$0.0121.5 min#62
Claude Haiku 5.5 (medium)15911537 to 163676.7$0.00670.8 min#74
Claude Haiku 5.5 (low)14771418 to 152275.2$0.00450.5 min#86

Cost is one script at the API list price of the route used, uncached and without judging. Time is the average for one accepted script, retries included.

Computed Findings Across Thinking Levels

Each finding is computed from the configurations above. The 95% ranges describe individual configurations; overlap alone establishes neither a tie nor an order for a pair.

Writing Elo at Each Shared Level Name

LevelClaude Opus 5Claude Haiku 5.5Gap95% Ranges
high21841686+498Separate
max22871966+321Separate

Gap is Claude Opus 5 minus Claude Haiku 5.5 at the same level name, using each model's highest-rated configuration at that level.

Two Chosen Configurations Compared

Pick one configuration of each model to compare their metric scores, consistency across drafts and per-script scores. Each menu starts on the model's best configuration.

Scores by Metric

Claude Opus 5 (max effort)Claude Haiku 5.5 (max effort)Solid bar: higher score in that row

Claude Opus 5 (max effort) scores higher on five of the nine metrics and Claude Haiku 5.5 (max effort) on four. Weighted by each metric's share of the Writing Score, Tone & Voice Match does the most to put Claude Opus 5 (max effort) ahead, with a 4.4-point lead on that metric. The weight beside each metric is its share of the Writing Score.

Tone & Voice Match19% weight
80.8
76.4
Writing Craft & Clarity13% weight
80.5
77.8
Substance, Accuracy & Value15% weight
84.5
85.6
Continuity & Emotion14% weight
83.7
78.7
YouTube Best Practices12% weight
83.5
76.8
Hook Strength10% weight
82.3
78.2
Length Adherence8% weight
92.0
97.3
Slop Score5% weight
85.1
86.0
Visual Cue Quality4% weight
79.8
80.9

Consistency Across Drafts

Claude Opus 5 (max effort)Claude Haiku 5.5 (max effort)
Score spread across drafts (±)1.101.55
Lowest script score81.9 Script 1078.2 Script 10
Highest script score84.5 Script 782.6 Script 4
Range across scripts2.64.4

Spread is how much the Writing Score moves between a configuration's drafts, and lower is steadier. Green marks the better number in each row.

Scores by Script

Writing Score on each reference script. Gap is Claude Opus 5 (max effort) minus Claude Haiku 5.5 (max effort), and the higher score is in bold.

ScriptClaude Opus 5 (max effort)Claude Haiku 5.5 (max effort)Gap
Script 1 opinion / warning explainer83.080.3+2.7
Script 2 news-analysis / skeptical explainer83.480.7+2.7
Script 3 personal roadmap / opinion83.280.8+2.4
Script 4 short explainer83.282.6+0.6
Script 5 news-analysis / opinion explainer83.381.7+1.6
Script 6 founder announcement / personal origin story83.481.1+2.3
Script 7 personal technical walkthrough / agentic workflow case study84.580.0+4.5
Script 8 product release announcement / personal observation83.081.3+1.7
Script 9 career guide / hiring analysis83.681.5+2.1
Script 10 engineering process walkthrough / presentation adaptation81.978.2+3.7

Compare With Another Model

Pick a second model. A pair with its own written match-up opens that page, and any other pair opens in the comparison tool.

Nearby Match-Ups by Best Writing Elo

Every configuration wrote five drafts for each of the same 10 briefs, from the same prompt. Generation was closed-book, with the prompt as the only source and no web search or other tools. Each judge scores every draft once: Claude Opus 5.5 at medium effort through the Claude Code subscription, GPT-6.1 Sol at medium reasoning through the Codex CLI, GLM 5.3 Flash through OpenRouter, and Jev 1.13, a typed judge from TypeSafe that answers each metric as a multiple-choice question. The channel's finished script is their reference for the voice, and it gets no score on the board itself.

← All comparisons